WEBVTT Kind: captions Language: en 00:00:13.880 --> 00:00:15.789 align:start position:0% hello<00:00:14.120> everybody<00:00:14.719> I'm<00:00:14.879> Alexi<00:00:15.240> kov<00:00:15.719> the 00:00:15.789 --> 00:00:15.799 align:start position:0% hello everybody I'm Alexi kov the 00:00:15.799 --> 00:00:18.470 align:start position:0% hello everybody I'm Alexi kov the organizer<00:00:16.560> of<00:00:16.760> SF<00:00:17.160> text<00:00:17.480> a<00:00:17.600> new<00:00:17.760> metup<00:00:18.160> devoted 00:00:18.470 --> 00:00:18.480 align:start position:0% organizer of SF text a new metup devoted 00:00:18.480 --> 00:00:21.509 align:start position:0% organizer of SF text a new metup devoted to<00:00:18.680> text<00:00:18.880> mining<00:00:19.400> AI<00:00:19.880> search<00:00:20.720> NLP<00:00:21.320> and 00:00:21.509 --> 00:00:21.519 align:start position:0% to text mining AI search NLP and 00:00:21.519 --> 00:00:24.029 align:start position:0% to text mining AI search NLP and basically<00:00:22.240> discovering<00:00:22.560> human<00:00:22.840> anten<00:00:23.480> behind 00:00:24.029 --> 00:00:24.039 align:start position:0% basically discovering human anten behind 00:00:24.039 --> 00:00:27.070 align:start position:0% basically discovering human anten behind you<00:00:24.160> know<00:00:24.400> the<00:00:24.960> strings<00:00:25.359> of<00:00:25.599> text<00:00:26.599> and<00:00:26.840> this<00:00:26.960> is 00:00:27.070 --> 00:00:27.080 align:start position:0% you know the strings of text and this is 00:00:27.080 --> 00:00:29.630 align:start position:0% you know the strings of text and this is our<00:00:27.279> second<00:00:27.519> meet<00:00:27.760> up<00:00:27.920> we're<00:00:28.160> here<00:00:28.359> at<00:00:28.640> lithium 00:00:29.630 --> 00:00:29.640 align:start position:0% our second meet up we're here at lithium 00:00:29.640 --> 00:00:32.830 align:start position:0% our second meet up we're here at lithium and<00:00:30.080> have<00:00:30.320> ad<00:00:31.320> uh<00:00:31.439> lead<00:00:31.759> research<00:00:32.200> engineer<00:00:32.719> at 00:00:32.830 --> 00:00:32.840 align:start position:0% and have ad uh lead research engineer at 00:00:32.840 --> 00:00:35.470 align:start position:0% and have ad uh lead research engineer at lithium<00:00:33.280> with<00:00:33.440> us<00:00:33.719> and<00:00:34.520> uh<00:00:34.719> we<00:00:34.920> formerly<00:00:35.360> were 00:00:35.470 --> 00:00:35.480 align:start position:0% lithium with us and uh we formerly were 00:00:35.480 --> 00:00:37.709 align:start position:0% lithium with us and uh we formerly were colleagues<00:00:35.760> at<00:00:35.920> Cloud<00:00:36.640> so<00:00:37.000> I<00:00:37.160> know<00:00:37.440> quite<00:00:37.600> a 00:00:37.709 --> 00:00:37.719 align:start position:0% colleagues at Cloud so I know quite a 00:00:37.719 --> 00:00:40.510 align:start position:0% colleagues at Cloud so I know quite a bit<00:00:37.960> about<00:00:38.399> cloud<00:00:39.399> and<00:00:39.760> uh<00:00:39.879> the<00:00:40.039> topic<00:00:40.280> today 00:00:40.510 --> 00:00:40.520 align:start position:0% bit about cloud and uh the topic today 00:00:40.520 --> 00:00:43.110 align:start position:0% bit about cloud and uh the topic today is<00:00:40.680> topics<00:00:41.640> U<00:00:41.960> which<00:00:42.079> is<00:00:42.239> something<00:00:42.960> that 00:00:43.110 --> 00:00:43.120 align:start position:0% is topics U which is something that 00:00:43.120 --> 00:00:45.510 align:start position:0% is topics U which is something that cloud<00:00:43.440> does<00:00:43.719> really<00:00:44.000> well<00:00:44.200> and<00:00:44.360> Adia<00:00:45.039> works<00:00:45.320> on 00:00:45.510 --> 00:00:45.520 align:start position:0% cloud does really well and Adia works on 00:00:45.520 --> 00:00:47.670 align:start position:0% cloud does really well and Adia works on is<00:00:45.640> going<00:00:45.760> to<00:00:45.960> talk<00:00:46.160> about<00:00:46.960> tonight<00:00:47.320> so<00:00:47.440> we'll 00:00:47.670 --> 00:00:47.680 align:start position:0% is going to talk about tonight so we'll 00:00:47.680 --> 00:00:50.069 align:start position:0% is going to talk about tonight so we'll have<00:00:47.840> a<00:00:48.199> the<00:00:48.360> talk<00:00:48.600> separately<00:00:49.559> so<00:00:49.840> I'm<00:00:49.960> just 00:00:50.069 --> 00:00:50.079 align:start position:0% have a the talk separately so I'm just 00:00:50.079 --> 00:00:51.470 align:start position:0% have a the talk separately so I'm just going<00:00:50.160> to<00:00:50.320> ask<00:00:50.440> a<00:00:50.520> few<00:00:50.719> questions<00:00:51.039> to<00:00:51.239> kind<00:00:51.360> of 00:00:51.470 --> 00:00:51.480 align:start position:0% going to ask a few questions to kind of 00:00:51.480 --> 00:00:53.549 align:start position:0% going to ask a few questions to kind of get<00:00:51.640> more<00:00:51.879> perspective<00:00:52.399> on<00:00:52.640> on<00:00:52.920> this<00:00:53.280> whole 00:00:53.549 --> 00:00:53.559 align:start position:0% get more perspective on on this whole 00:00:53.559 --> 00:00:55.910 align:start position:0% get more perspective on on this whole Space<00:00:54.359> yeah<00:00:55.000> so<00:00:55.199> it's<00:00:55.359> great<00:00:55.520> to<00:00:55.600> have<00:00:55.719> your 00:00:55.910 --> 00:00:55.920 align:start position:0% Space yeah so it's great to have your 00:00:55.920 --> 00:01:00.389 align:start position:0% Space yeah so it's great to have your DET<00:00:56.199> with<00:00:56.320> us<00:00:56.840> thanks<00:00:57.039> foring<00:00:57.440> us<00:00:58.440> uh<00:00:59.199> so<00:00:59.760> uh<00:01:00.120> uh 00:01:00.389 --> 00:01:00.399 align:start position:0% DET with us thanks foring us uh so uh uh 00:01:00.399 --> 00:01:02.549 align:start position:0% DET with us thanks foring us uh so uh uh how<00:01:00.760> did<00:01:00.960> you<00:01:01.440> get<00:01:01.640> interested<00:01:02.039> in<00:01:02.160> this<00:01:02.320> whole 00:01:02.549 --> 00:01:02.559 align:start position:0% how did you get interested in this whole 00:01:02.559 --> 00:01:04.750 align:start position:0% how did you get interested in this whole space<00:01:02.960> I<00:01:03.079> know<00:01:03.320> that<00:01:03.440> you<00:01:03.559> know<00:01:04.439> what<00:01:04.559> you<00:01:04.680> were 00:01:04.750 --> 00:01:04.760 align:start position:0% space I know that you know what you were 00:01:04.760 --> 00:01:06.550 align:start position:0% space I know that you know what you were doing<00:01:04.960> at<00:01:05.119> Cloud<00:01:05.680> you<00:01:05.799> know<00:01:06.240> and<00:01:06.360> you're 00:01:06.550 --> 00:01:06.560 align:start position:0% doing at Cloud you know and you're 00:01:06.560 --> 00:01:08.350 align:start position:0% doing at Cloud you know and you're really<00:01:06.760> interested<00:01:07.119> in<00:01:07.320> the<00:01:07.479> science<00:01:08.119> of 00:01:08.350 --> 00:01:08.360 align:start position:0% really interested in the science of 00:01:08.360 --> 00:01:10.149 align:start position:0% really interested in the science of these<00:01:08.560> things<00:01:08.840> how<00:01:09.000> they<00:01:09.200> work<00:01:09.759> so<00:01:09.920> can<00:01:10.040> you 00:01:10.149 --> 00:01:10.159 align:start position:0% these things how they work so can you 00:01:10.159 --> 00:01:11.469 align:start position:0% these things how they work so can you tell<00:01:10.400> me<00:01:10.560> a<00:01:10.680> little<00:01:10.840> bit<00:01:11.080> about<00:01:11.280> your 00:01:11.469 --> 00:01:11.479 align:start position:0% tell me a little bit about your 00:01:11.479 --> 00:01:14.030 align:start position:0% tell me a little bit about your background<00:01:12.280> you<00:01:12.400> know<00:01:12.799> and<00:01:13.119> what<00:01:13.280> brought<00:01:13.560> you 00:01:14.030 --> 00:01:14.040 align:start position:0% background you know and what brought you 00:01:14.040 --> 00:01:16.749 align:start position:0% background you know and what brought you to<00:01:14.640> to<00:01:14.840> these<00:01:15.080> problems<00:01:15.600> definitely<00:01:16.439> um<00:01:16.680> I 00:01:16.749 --> 00:01:16.759 align:start position:0% to to these problems definitely um I 00:01:16.759 --> 00:01:19.109 align:start position:0% to to these problems definitely um I think<00:01:16.920> it<00:01:17.080> started<00:01:17.479> mainly<00:01:17.960> when<00:01:18.280> I<00:01:18.400> started 00:01:19.109 --> 00:01:19.119 align:start position:0% think it started mainly when I started 00:01:19.119 --> 00:01:21.510 align:start position:0% think it started mainly when I started looking<00:01:19.439> at<00:01:19.960> uh<00:01:20.400> I<00:01:20.560> I<00:01:20.640> come<00:01:20.880> from<00:01:21.119> originally 00:01:21.510 --> 00:01:21.520 align:start position:0% looking at uh I I come from originally 00:01:21.520 --> 00:01:23.190 align:start position:0% looking at uh I I come from originally from<00:01:21.680> an<00:01:22.200> electrical<00:01:22.640> engineering 00:01:23.190 --> 00:01:23.200 align:start position:0% from an electrical engineering 00:01:23.200 --> 00:01:25.390 align:start position:0% from an electrical engineering background<00:01:24.119> and<00:01:24.280> I<00:01:24.400> was<00:01:24.600> looking<00:01:24.799> at<00:01:25.000> wireless 00:01:25.390 --> 00:01:25.400 align:start position:0% background and I was looking at wireless 00:01:25.400 --> 00:01:28.230 align:start position:0% background and I was looking at wireless networks<00:01:26.040> as<00:01:26.280> my<00:01:26.600> initial<00:01:27.240> field<00:01:27.479> of<00:01:27.640> study 00:01:28.230 --> 00:01:28.240 align:start position:0% networks as my initial field of study 00:01:28.240 --> 00:01:30.230 align:start position:0% networks as my initial field of study okay<00:01:28.680> uh<00:01:29.119> from<00:01:29.320> wireless<00:01:29.640> network<00:01:29.960> Works 00:01:30.230 --> 00:01:30.240 align:start position:0% okay uh from wireless network Works 00:01:30.240 --> 00:01:32.789 align:start position:0% okay uh from wireless network Works communication<00:01:31.040> was<00:01:31.280> a<00:01:31.880> uh<00:01:32.079> kind<00:01:32.240> of<00:01:32.360> an<00:01:32.520> easy 00:01:32.789 --> 00:01:32.799 align:start position:0% communication was a uh kind of an easy 00:01:32.799 --> 00:01:34.710 align:start position:0% communication was a uh kind of an easy walk<00:01:33.079> through<00:01:33.200> and<00:01:33.399> there<00:01:34.079> uh<00:01:34.240> information 00:01:34.710 --> 00:01:34.720 align:start position:0% walk through and there uh information 00:01:34.720 --> 00:01:36.429 align:start position:0% walk through and there uh information Theory<00:01:35.159> became<00:01:35.479> like<00:01:35.600> a<00:01:35.759> very<00:01:35.960> interesting 00:01:36.429 --> 00:01:36.439 align:start position:0% Theory became like a very interesting 00:01:36.439 --> 00:01:38.630 align:start position:0% Theory became like a very interesting subject<00:01:37.159> as<00:01:37.280> I<00:01:37.560> started<00:01:37.880> going<00:01:38.159> deeper<00:01:38.439> into 00:01:38.630 --> 00:01:38.640 align:start position:0% subject as I started going deeper into 00:01:38.640 --> 00:01:40.310 align:start position:0% subject as I started going deeper into that<00:01:38.799> whole<00:01:39.040> life<00:01:39.280> I<00:01:39.399> realized<00:01:39.799> that<00:01:39.960> it's 00:01:40.310 --> 00:01:40.320 align:start position:0% that whole life I realized that it's 00:01:40.320 --> 00:01:43.510 align:start position:0% that whole life I realized that it's probably<00:01:41.320> uh<00:01:41.880> easier<00:01:42.200> to<00:01:42.399> work<00:01:42.759> with<00:01:43.240> when<00:01:43.360> you 00:01:43.510 --> 00:01:43.520 align:start position:0% probably uh easier to work with when you 00:01:43.520 --> 00:01:45.630 align:start position:0% probably uh easier to work with when you look<00:01:43.680> at<00:01:43.840> it<00:01:44.000> from<00:01:44.280> the<00:01:44.439> higher<00:01:44.759> abstraction 00:01:45.630 --> 00:01:45.640 align:start position:0% look at it from the higher abstraction 00:01:45.640 --> 00:01:48.270 align:start position:0% look at it from the higher abstraction uh<00:01:45.799> and<00:01:46.399> that's<00:01:46.600> when<00:01:46.920> I<00:01:47.079> kind<00:01:47.240> of<00:01:47.560> moved<00:01:47.920> into 00:01:48.270 --> 00:01:48.280 align:start position:0% uh and that's when I kind of moved into 00:01:48.280 --> 00:01:50.190 align:start position:0% uh and that's when I kind of moved into the<00:01:48.799> domain<00:01:49.079> of<00:01:49.240> like<00:01:49.360> machine<00:01:49.640> learning<00:01:50.000> in 00:01:50.190 --> 00:01:50.200 align:start position:0% the domain of like machine learning in 00:01:50.200 --> 00:01:52.950 align:start position:0% the domain of like machine learning in particular<00:01:51.040> so<00:01:51.520> uh<00:01:51.719> at<00:01:52.000> I<00:01:52.119> did<00:01:52.280> my<00:01:52.479> masters 00:01:52.950 --> 00:01:52.960 align:start position:0% particular so uh at I did my masters 00:01:52.960 --> 00:01:55.230 align:start position:0% particular so uh at I did my masters from<00:01:53.159> Stanford<00:01:53.920> uh<00:01:54.200> and<00:01:54.439> there<00:01:54.719> I<00:01:54.880> mainly 00:01:55.230 --> 00:01:55.240 align:start position:0% from Stanford uh and there I mainly 00:01:55.240 --> 00:01:56.830 align:start position:0% from Stanford uh and there I mainly specialized<00:01:55.759> in<00:01:55.880> machine<00:01:56.119> learning<00:01:56.439> and<00:01:56.600> data 00:01:56.830 --> 00:01:56.840 align:start position:0% specialized in machine learning and data 00:01:56.840 --> 00:01:58.990 align:start position:0% specialized in machine learning and data mining<00:01:57.600> and<00:01:57.719> that<00:01:57.880> really<00:01:58.119> got<00:01:58.280> me<00:01:58.479> interested 00:01:58.990 --> 00:01:59.000 align:start position:0% mining and that really got me interested 00:01:59.000 --> 00:02:00.910 align:start position:0% mining and that really got me interested into<00:01:59.280> the<00:01:59.399> various<00:02:00.039> domains<00:02:00.520> where<00:02:00.680> you<00:02:00.759> can 00:02:00.910 --> 00:02:00.920 align:start position:0% into the various domains where you can 00:02:00.920 --> 00:02:03.789 align:start position:0% into the various domains where you can apply<00:02:01.200> machine<00:02:01.479> learning<00:02:02.479> and<00:02:02.920> uh<00:02:03.240> text<00:02:03.640> was 00:02:03.789 --> 00:02:03.799 align:start position:0% apply machine learning and uh text was 00:02:03.799 --> 00:02:06.350 align:start position:0% apply machine learning and uh text was one<00:02:03.920> of<00:02:04.039> the<00:02:04.159> big<00:02:04.320> ones<00:02:04.799> and<00:02:05.320> uh<00:02:06.079> that<00:02:06.200> was 00:02:06.350 --> 00:02:06.360 align:start position:0% one of the big ones and uh that was 00:02:06.360 --> 00:02:08.830 align:start position:0% one of the big ones and uh that was around<00:02:06.640> maybe<00:02:06.840> four<00:02:07.079> years<00:02:07.280> ago<00:02:07.640> when<00:02:08.560> uh<00:02:08.720> you 00:02:08.830 --> 00:02:08.840 align:start position:0% around maybe four years ago when uh you 00:02:08.840 --> 00:02:10.949 align:start position:0% around maybe four years ago when uh you know<00:02:09.000> the<00:02:09.360> social<00:02:09.640> media<00:02:09.920> boom<00:02:10.239> was<00:02:10.440> just<00:02:10.720> like 00:02:10.949 --> 00:02:10.959 align:start position:0% know the social media boom was just like 00:02:10.959 --> 00:02:13.350 align:start position:0% know the social media boom was just like taking<00:02:11.280> off<00:02:11.640> there<00:02:11.840> A<00:02:11.959> lot<00:02:12.120> of<00:02:12.640> lot<00:02:12.800> of<00:02:13.000> people 00:02:13.350 --> 00:02:13.360 align:start position:0% taking off there A lot of lot of people 00:02:13.360 --> 00:02:15.589 align:start position:0% taking off there A lot of lot of people everybody<00:02:14.040> uh<00:02:14.599> getting<00:02:14.840> overwhelmed<00:02:15.360> with<00:02:15.480> so 00:02:15.589 --> 00:02:15.599 align:start position:0% everybody uh getting overwhelmed with so 00:02:15.599 --> 00:02:17.550 align:start position:0% everybody uh getting overwhelmed with so much<00:02:15.879> data<00:02:16.160> around<00:02:16.519> and<00:02:16.959> I<00:02:17.040> thought<00:02:17.160> it<00:02:17.280> was<00:02:17.400> a 00:02:17.550 --> 00:02:17.560 align:start position:0% much data around and I thought it was a 00:02:17.560 --> 00:02:19.430 align:start position:0% much data around and I thought it was a per<00:02:17.800> perfect<00:02:18.120> opportunity<00:02:18.640> to<00:02:18.840> kind<00:02:19.000> of<00:02:19.239> look 00:02:19.430 --> 00:02:19.440 align:start position:0% per perfect opportunity to kind of look 00:02:19.440 --> 00:02:22.790 align:start position:0% per perfect opportunity to kind of look at<00:02:19.920> uh<00:02:20.440> uh<00:02:20.640> information<00:02:21.040> in<00:02:21.239> social<00:02:21.519> media<00:02:22.440> and 00:02:22.790 --> 00:02:22.800 align:start position:0% at uh uh information in social media and 00:02:22.800 --> 00:02:24.390 align:start position:0% at uh uh information in social media and uh<00:02:23.000> how<00:02:23.160> machine<00:02:23.440> learning<00:02:23.760> can<00:02:23.920> be<00:02:24.080> applied 00:02:24.390 --> 00:02:24.400 align:start position:0% uh how machine learning can be applied 00:02:24.400 --> 00:02:27.309 align:start position:0% uh how machine learning can be applied to<00:02:24.599> that<00:02:25.160> aspect<00:02:25.680> and<00:02:26.080> uh<00:02:26.640> yeah<00:02:26.760> so<00:02:26.920> I<00:02:27.040> joined 00:02:27.309 --> 00:02:27.319 align:start position:0% to that aspect and uh yeah so I joined 00:02:27.319 --> 00:02:30.509 align:start position:0% to that aspect and uh yeah so I joined Cloud<00:02:28.040> at<00:02:28.160> that<00:02:28.360> point<00:02:29.040> and<00:02:29.480> with<00:02:30.120> basically 00:02:30.509 --> 00:02:30.519 align:start position:0% Cloud at that point and with basically 00:02:30.519 --> 00:02:33.150 align:start position:0% Cloud at that point and with basically there<00:02:30.640> was<00:02:30.879> this<00:02:31.160> huge<00:02:31.680> data<00:02:32.040> set<00:02:32.319> to<00:02:32.599> work<00:02:32.840> on 00:02:33.150 --> 00:02:33.160 align:start position:0% there was this huge data set to work on 00:02:33.160 --> 00:02:35.869 align:start position:0% there was this huge data set to work on and<00:02:33.840> uh<00:02:33.959> it's<00:02:34.120> been<00:02:34.319> pretty<00:02:34.560> exciting<00:02:35.360> yes 00:02:35.869 --> 00:02:35.879 align:start position:0% and uh it's been pretty exciting yes 00:02:35.879 --> 00:02:37.589 align:start position:0% and uh it's been pretty exciting yes yeah<00:02:36.000> I<00:02:36.080> know<00:02:36.440> you<00:02:36.560> know<00:02:36.840> some<00:02:37.080> of<00:02:37.239> the<00:02:37.319> story 00:02:37.589 --> 00:02:37.599 align:start position:0% yeah I know you know some of the story 00:02:37.599 --> 00:02:39.309 align:start position:0% yeah I know you know some of the story and<00:02:37.720> that's<00:02:37.879> you<00:02:38.000> know<00:02:38.120> it<00:02:38.480> always<00:02:38.720> amazes<00:02:39.120> me 00:02:39.309 --> 00:02:39.319 align:start position:0% and that's you know it always amazes me 00:02:39.319 --> 00:02:40.949 align:start position:0% and that's you know it always amazes me that<00:02:39.680> you<00:02:39.800> know<00:02:39.959> with<00:02:40.080> a<00:02:40.239> relatively<00:02:40.680> small 00:02:40.949 --> 00:02:40.959 align:start position:0% that you know with a relatively small 00:02:40.959 --> 00:02:43.309 align:start position:0% that you know with a relatively small team<00:02:41.720> Cloud<00:02:42.040> was<00:02:42.200> able<00:02:42.400> to<00:02:42.560> achieve<00:02:42.879> so<00:02:43.040> much 00:02:43.309 --> 00:02:43.319 align:start position:0% team Cloud was able to achieve so much 00:02:43.319 --> 00:02:45.790 align:start position:0% team Cloud was able to achieve so much right<00:02:43.720> because<00:02:43.879> so<00:02:44.080> now<00:02:44.640> as<00:02:44.760> I<00:02:45.159> understand<00:02:45.560> it 00:02:45.790 --> 00:02:45.800 align:start position:0% right because so now as I understand it 00:02:45.800 --> 00:02:47.589 align:start position:0% right because so now as I understand it it<00:02:46.040> ingests<00:02:46.640> essentially<00:02:47.080> most<00:02:47.319> of<00:02:47.480> the 00:02:47.589 --> 00:02:47.599 align:start position:0% it ingests essentially most of the 00:02:47.599 --> 00:02:50.550 align:start position:0% it ingests essentially most of the Social<00:02:47.920> Web<00:02:48.480> yeah<00:02:49.040> a<00:02:49.159> lot<00:02:49.280> of<00:02:49.440> it<00:02:49.680> yes<00:02:50.080> I<00:02:50.159> mean 00:02:50.550 --> 00:02:50.560 align:start position:0% Social Web yeah a lot of it yes I mean 00:02:50.560 --> 00:02:52.390 align:start position:0% Social Web yeah a lot of it yes I mean uh<00:02:50.680> I<00:02:50.760> mean<00:02:50.879> I<00:02:51.040> remember<00:02:51.440> you<00:02:51.519> know<00:02:51.760> it<00:02:51.920> was 00:02:52.390 --> 00:02:52.400 align:start position:0% uh I mean I remember you know it was 00:02:52.400 --> 00:02:54.470 align:start position:0% uh I mean I remember you know it was quite<00:02:52.560> a<00:02:52.680> few<00:02:52.840> itations<00:02:53.319> to<00:02:53.440> get<00:02:53.680> there<00:02:53.879> but 00:02:54.470 --> 00:02:54.480 align:start position:0% quite a few itations to get there but 00:02:54.480 --> 00:02:56.550 align:start position:0% quite a few itations to get there but how<00:02:55.080> I<00:02:55.159> mean<00:02:55.400> how<00:02:55.519> do<00:02:55.680> you<00:02:55.800> find<00:02:56.120> balance 00:02:56.550 --> 00:02:56.560 align:start position:0% how I mean how do you find balance 00:02:56.560 --> 00:02:58.390 align:start position:0% how I mean how do you find balance basically<00:02:57.040> with<00:02:57.239> you<00:02:57.360> know<00:02:58.040> kind<00:02:58.159> of<00:02:58.239> a 00:02:58.390 --> 00:02:58.400 align:start position:0% basically with you know kind of a 00:02:58.400 --> 00:03:01.750 align:start position:0% basically with you know kind of a startup<00:02:58.879> setting<00:02:59.400> you<00:02:59.519> know<00:02:59.920> be<00:03:00.040> able<00:03:00.239> to<00:03:00.800> to 00:03:01.750 --> 00:03:01.760 align:start position:0% startup setting you know be able to to 00:03:01.760 --> 00:03:04.630 align:start position:0% startup setting you know be able to to regularly<00:03:02.519> daily<00:03:03.239> you<00:03:03.360> know<00:03:03.879> ingest<00:03:04.319> all<00:03:04.480> of 00:03:04.630 --> 00:03:04.640 align:start position:0% regularly daily you know ingest all of 00:03:04.640 --> 00:03:06.589 align:start position:0% regularly daily you know ingest all of this<00:03:04.840> data<00:03:05.200> and<00:03:05.480> do<00:03:05.720> some<00:03:05.920> science<00:03:06.239> on<00:03:06.400> this 00:03:06.589 --> 00:03:06.599 align:start position:0% this data and do some science on this 00:03:06.599 --> 00:03:08.309 align:start position:0% this data and do some science on this like<00:03:06.760> how<00:03:06.959> how<00:03:07.080> do<00:03:07.159> you<00:03:07.360> balance<00:03:08.159> the 00:03:08.309 --> 00:03:08.319 align:start position:0% like how how do you balance the 00:03:08.319 --> 00:03:10.430 align:start position:0% like how how do you balance the engineering<00:03:09.080> work<00:03:09.480> versus<00:03:09.840> the<00:03:09.959> research 00:03:10.430 --> 00:03:10.440 align:start position:0% engineering work versus the research 00:03:10.440 --> 00:03:13.550 align:start position:0% engineering work versus the research work<00:03:10.959> right<00:03:11.319> right<00:03:11.720> so<00:03:12.040> I<00:03:12.159> think<00:03:12.640> uh<00:03:13.319> I<00:03:13.400> think 00:03:13.550 --> 00:03:13.560 align:start position:0% work right right so I think uh I think 00:03:13.560 --> 00:03:15.350 align:start position:0% work right right so I think uh I think the<00:03:13.760> important<00:03:14.280> part<00:03:14.519> is<00:03:14.680> the<00:03:14.799> engineering 00:03:15.350 --> 00:03:15.360 align:start position:0% the important part is the engineering 00:03:15.360 --> 00:03:17.270 align:start position:0% the important part is the engineering work<00:03:15.920> at<00:03:16.000> the<00:03:16.120> end<00:03:16.239> of<00:03:16.319> the<00:03:16.480> day<00:03:16.680> because 00:03:17.270 --> 00:03:17.280 align:start position:0% work at the end of the day because 00:03:17.280 --> 00:03:19.789 align:start position:0% work at the end of the day because unless<00:03:17.560> you're<00:03:17.799> able<00:03:18.000> to<00:03:18.239> consume<00:03:19.200> uh<00:03:19.599> this 00:03:19.789 --> 00:03:19.799 align:start position:0% unless you're able to consume uh this 00:03:19.799 --> 00:03:21.670 align:start position:0% unless you're able to consume uh this kind<00:03:19.959> of<00:03:20.120> data<00:03:20.519> and<00:03:20.680> build<00:03:20.920> up<00:03:21.120> Frameworks 00:03:21.670 --> 00:03:21.680 align:start position:0% kind of data and build up Frameworks 00:03:21.680 --> 00:03:23.430 align:start position:0% kind of data and build up Frameworks that<00:03:21.799> allow<00:03:22.080> you<00:03:22.159> to<00:03:22.400> kind<00:03:22.560> of<00:03:23.000> iterate 00:03:23.430 --> 00:03:23.440 align:start position:0% that allow you to kind of iterate 00:03:23.440 --> 00:03:25.990 align:start position:0% that allow you to kind of iterate quickly<00:03:24.040> or<00:03:24.239> try<00:03:24.400> to<00:03:24.560> do<00:03:24.760> things<00:03:25.120> which<00:03:25.560> where 00:03:25.990 --> 00:03:26.000 align:start position:0% quickly or try to do things which where 00:03:26.000 --> 00:03:28.030 align:start position:0% quickly or try to do things which where you<00:03:26.120> are<00:03:26.280> able<00:03:26.480> to<00:03:26.920> push<00:03:27.200> things<00:03:27.360> out<00:03:27.599> fast<00:03:27.879> to 00:03:28.030 --> 00:03:28.040 align:start position:0% you are able to push things out fast to 00:03:28.040 --> 00:03:29.990 align:start position:0% you are able to push things out fast to production<00:03:28.599> and<00:03:28.720> so<00:03:28.879> on<00:03:29.280> no<00:03:29.439> matter<00:03:29.799> how 00:03:29.990 --> 00:03:30.000 align:start position:0% production and so on no matter how 00:03:30.000 --> 00:03:31.670 align:start position:0% production and so on no matter how complex<00:03:30.360> your<00:03:30.560> algorithms<00:03:31.200> are<00:03:31.400> it's<00:03:31.560> not 00:03:31.670 --> 00:03:31.680 align:start position:0% complex your algorithms are it's not 00:03:31.680 --> 00:03:32.949 align:start position:0% complex your algorithms are it's not going<00:03:31.799> to<00:03:31.959> amount<00:03:32.200> to<00:03:32.319> anything<00:03:32.640> at<00:03:32.760> the<00:03:32.840> end 00:03:32.949 --> 00:03:32.959 align:start position:0% going to amount to anything at the end 00:03:32.959 --> 00:03:35.509 align:start position:0% going to amount to anything at the end of<00:03:33.080> the<00:03:33.200> day<00:03:33.400> so<00:03:33.920> I<00:03:34.040> think<00:03:34.519> uh<00:03:34.840> that<00:03:34.959> is<00:03:35.200> where 00:03:35.509 --> 00:03:35.519 align:start position:0% of the day so I think uh that is where 00:03:35.519 --> 00:03:37.830 align:start position:0% of the day so I think uh that is where like<00:03:35.720> a<00:03:35.840> lot<00:03:36.000> of<00:03:36.680> effort<00:03:37.080> went<00:03:37.319> in<00:03:37.599> in<00:03:37.760> the 00:03:37.830 --> 00:03:37.840 align:start position:0% like a lot of effort went in in the 00:03:37.840 --> 00:03:40.710 align:start position:0% like a lot of effort went in in the beginning<00:03:38.159> years<00:03:38.360> of<00:03:38.519> cloud<00:03:38.879> at<00:03:39.000> least<00:03:39.239> and<00:03:40.080> uh 00:03:40.710 --> 00:03:40.720 align:start position:0% beginning years of cloud at least and uh 00:03:40.720 --> 00:03:42.429 align:start position:0% beginning years of cloud at least and uh finally<00:03:41.080> when<00:03:41.280> we<00:03:41.480> did<00:03:41.720> have<00:03:41.840> a<00:03:41.959> framework 00:03:42.429 --> 00:03:42.439 align:start position:0% finally when we did have a framework 00:03:42.439 --> 00:03:44.229 align:start position:0% finally when we did have a framework where<00:03:42.599> we<00:03:42.720> could<00:03:42.879> ingest<00:03:43.239> this<00:03:43.439> data<00:03:43.799> and<00:03:43.920> run 00:03:44.229 --> 00:03:44.239 align:start position:0% where we could ingest this data and run 00:03:44.239 --> 00:03:46.550 align:start position:0% where we could ingest this data and run like<00:03:44.840> pipelines<00:03:45.319> every<00:03:45.519> day<00:03:45.959> then<00:03:46.159> it<00:03:46.319> became 00:03:46.550 --> 00:03:46.560 align:start position:0% like pipelines every day then it became 00:03:46.560 --> 00:03:48.750 align:start position:0% like pipelines every day then it became easy<00:03:46.799> to<00:03:47.000> kind<00:03:47.120> of<00:03:47.319> let's<00:03:47.480> say<00:03:47.879> let's<00:03:48.120> try<00:03:48.599> try 00:03:48.750 --> 00:03:48.760 align:start position:0% easy to kind of let's say let's try try 00:03:48.760 --> 00:03:50.589 align:start position:0% easy to kind of let's say let's try try out<00:03:48.959> this<00:03:49.120> algorithm<00:03:49.680> let's<00:03:49.840> try<00:03:50.040> out<00:03:50.480> this 00:03:50.589 --> 00:03:50.599 align:start position:0% out this algorithm let's try out this 00:03:50.599 --> 00:03:52.550 align:start position:0% out this algorithm let's try out this model<00:03:50.959> let's<00:03:51.120> try<00:03:51.280> out<00:03:51.480> these<00:03:51.640> features<00:03:52.360> let's 00:03:52.550 --> 00:03:52.560 align:start position:0% model let's try out these features let's 00:03:52.560 --> 00:03:55.270 align:start position:0% model let's try out these features let's maybe<00:03:52.840> vary<00:03:53.120> it<00:03:53.239> up<00:03:53.439> a<00:03:53.560> bit<00:03:53.840> and<00:03:53.920> then<00:03:54.280> maybe 00:03:55.270 --> 00:03:55.280 align:start position:0% maybe vary it up a bit and then maybe 00:03:55.280 --> 00:03:57.470 align:start position:0% maybe vary it up a bit and then maybe the<00:03:55.400> Cycles<00:03:55.840> became<00:03:56.159> faster<00:03:56.599> and<00:03:57.000> iteration 00:03:57.470 --> 00:03:57.480 align:start position:0% the Cycles became faster and iteration 00:03:57.480 --> 00:04:00.030 align:start position:0% the Cycles became faster and iteration became<00:03:57.799> faster<00:03:58.120> so<00:03:58.319> I<00:03:58.480> guess<00:03:59.159> uh<00:03:59.319> engine<00:03:59.920> ing 00:04:00.030 --> 00:04:00.040 align:start position:0% became faster so I guess uh engine ing 00:04:00.040 --> 00:04:01.869 align:start position:0% became faster so I guess uh engine ing played<00:04:00.599> plays<00:04:00.840> a<00:04:00.959> very<00:04:01.159> big<00:04:01.360> role<00:04:01.599> when<00:04:01.720> it 00:04:01.869 --> 00:04:01.879 align:start position:0% played plays a very big role when it 00:04:01.879 --> 00:04:03.750 align:start position:0% played plays a very big role when it comes<00:04:02.040> to<00:04:02.200> data<00:04:02.439> science<00:04:02.799> although<00:04:03.560> uh<00:04:03.640> when 00:04:03.750 --> 00:04:03.760 align:start position:0% comes to data science although uh when 00:04:03.760 --> 00:04:05.390 align:start position:0% comes to data science although uh when we<00:04:03.879> talk<00:04:04.079> about<00:04:04.280> data<00:04:04.519> science<00:04:04.799> in<00:04:04.959> particular 00:04:05.390 --> 00:04:05.400 align:start position:0% we talk about data science in particular 00:04:05.400 --> 00:04:08.949 align:start position:0% we talk about data science in particular we<00:04:05.720> often<00:04:06.040> ignore<00:04:06.560> that<00:04:06.920> aspect<00:04:07.239> of<00:04:07.400> it<00:04:07.959> yes<00:04:08.760> uh 00:04:08.949 --> 00:04:08.959 align:start position:0% we often ignore that aspect of it yes uh 00:04:08.959 --> 00:04:11.309 align:start position:0% we often ignore that aspect of it yes uh so<00:04:09.159> I<00:04:09.319> remember<00:04:10.200> uh<00:04:10.560> you<00:04:10.720> used<00:04:10.959> a<00:04:11.079> lot<00:04:11.200> of 00:04:11.309 --> 00:04:11.319 align:start position:0% so I remember uh you used a lot of 00:04:11.319 --> 00:04:13.270 align:start position:0% so I remember uh you used a lot of different<00:04:11.560> tools<00:04:12.000> you<00:04:12.120> know<00:04:12.480> we<00:04:12.680> at<00:04:12.840> Cloud<00:04:13.159> can 00:04:13.270 --> 00:04:13.280 align:start position:0% different tools you know we at Cloud can 00:04:13.280 --> 00:04:15.069 align:start position:0% different tools you know we at Cloud can you<00:04:13.400> talk<00:04:13.560> a<00:04:13.640> little<00:04:13.799> bit<00:04:14.040> about<00:04:14.400> what<00:04:14.599> kind<00:04:14.760> of 00:04:15.069 --> 00:04:15.079 align:start position:0% you talk a little bit about what kind of 00:04:15.079 --> 00:04:17.670 align:start position:0% you talk a little bit about what kind of uh<00:04:15.159> tools<00:04:15.400> are<00:04:15.600> using<00:04:16.600> uh<00:04:16.720> in<00:04:16.840> your<00:04:17.000> daily<00:04:17.280> work 00:04:17.670 --> 00:04:17.680 align:start position:0% uh tools are using uh in your daily work 00:04:17.680 --> 00:04:20.150 align:start position:0% uh tools are using uh in your daily work sure<00:04:18.359> yeah<00:04:18.560> so<00:04:19.040> uh<00:04:19.400> I<00:04:19.479> think<00:04:19.680> most<00:04:19.880> of<00:04:20.000> the 00:04:20.150 --> 00:04:20.160 align:start position:0% sure yeah so uh I think most of the 00:04:20.160 --> 00:04:21.590 align:start position:0% sure yeah so uh I think most of the stand<00:04:20.479> for<00:04:20.600> machine<00:04:20.880> learning<00:04:21.199> at<00:04:21.320> least<00:04:21.479> we 00:04:21.590 --> 00:04:21.600 align:start position:0% stand for machine learning at least we 00:04:21.600 --> 00:04:23.790 align:start position:0% stand for machine learning at least we use<00:04:21.840> most<00:04:21.959> of<00:04:22.079> the<00:04:22.160> standard<00:04:22.560> ones<00:04:22.919> like<00:04:23.320> uh 00:04:23.790 --> 00:04:23.800 align:start position:0% use most of the standard ones like uh 00:04:23.800 --> 00:04:27.390 align:start position:0% use most of the standard ones like uh python<00:04:24.160> libraries<00:04:24.759> and<00:04:25.440> uh<00:04:25.600> WCA<00:04:26.120> and<00:04:26.560> so<00:04:26.720> on<00:04:27.280> uh 00:04:27.390 --> 00:04:27.400 align:start position:0% python libraries and uh WCA and so on uh 00:04:27.400 --> 00:04:29.350 align:start position:0% python libraries and uh WCA and so on uh but<00:04:27.520> in<00:04:27.639> terms<00:04:27.880> of<00:04:28.000> the<00:04:28.120> data<00:04:28.400> pipeline<00:04:28.800> we<00:04:28.919> use 00:04:29.350 --> 00:04:29.360 align:start position:0% but in terms of the data pipeline we use 00:04:29.360 --> 00:04:32.430 align:start position:0% but in terms of the data pipeline we use uh<00:04:29.680> main<00:04:29.960> the<00:04:30.080> map<00:04:30.320> produ<00:04:30.639> stack<00:04:31.320> with<00:04:31.680> the 00:04:32.430 --> 00:04:32.440 align:start position:0% uh main the map produ stack with the 00:04:32.440 --> 00:04:35.390 align:start position:0% uh main the map produ stack with the with<00:04:32.560> our<00:04:33.000> hdfs<00:04:33.759> And<00:04:34.479> Hive<00:04:34.840> on<00:04:35.000> top<00:04:35.160> of<00:04:35.280> that 00:04:35.390 --> 00:04:35.400 align:start position:0% with our hdfs And Hive on top of that 00:04:35.400 --> 00:04:37.430 align:start position:0% with our hdfs And Hive on top of that for<00:04:35.560> our<00:04:35.720> quering<00:04:36.080> language<00:04:36.960> Hive<00:04:37.240> is 00:04:37.430 --> 00:04:37.440 align:start position:0% for our quering language Hive is 00:04:37.440 --> 00:04:38.870 align:start position:0% for our quering language Hive is actually<00:04:37.759> a<00:04:37.840> pretty<00:04:38.039> powerful<00:04:38.360> abstraction 00:04:38.870 --> 00:04:38.880 align:start position:0% actually a pretty powerful abstraction 00:04:38.880 --> 00:04:40.550 align:start position:0% actually a pretty powerful abstraction because<00:04:39.080> it<00:04:39.199> allows<00:04:39.520> us<00:04:39.680> to<00:04:40.160> write<00:04:40.400> like 00:04:40.550 --> 00:04:40.560 align:start position:0% because it allows us to write like 00:04:40.560 --> 00:04:42.070 align:start position:0% because it allows us to write like complex<00:04:40.919> queries<00:04:41.360> which<00:04:41.560> trigger<00:04:41.960> like 00:04:42.070 --> 00:04:42.080 align:start position:0% complex queries which trigger like 00:04:42.080 --> 00:04:43.670 align:start position:0% complex queries which trigger like multiple<00:04:42.479> jobs<00:04:42.840> without<00:04:43.120> actually<00:04:43.320> worrying 00:04:43.670 --> 00:04:43.680 align:start position:0% multiple jobs without actually worrying 00:04:43.680 --> 00:04:45.310 align:start position:0% multiple jobs without actually worrying about<00:04:44.240> how<00:04:44.560> how<00:04:44.680> you<00:04:44.759> would<00:04:44.960> go<00:04:45.080> about 00:04:45.310 --> 00:04:45.320 align:start position:0% about how how you would go about 00:04:45.320 --> 00:04:47.590 align:start position:0% about how how you would go about optimizing<00:04:45.919> those<00:04:46.120> things<00:04:46.800> and<00:04:47.199> the<00:04:47.320> really 00:04:47.590 --> 00:04:47.600 align:start position:0% optimizing those things and the really 00:04:47.600 --> 00:04:49.070 align:start position:0% optimizing those things and the really interesting<00:04:48.039> part<00:04:48.199> is<00:04:48.360> also<00:04:48.560> that<00:04:48.680> it<00:04:48.800> allows 00:04:49.070 --> 00:04:49.080 align:start position:0% interesting part is also that it allows 00:04:49.080 --> 00:04:51.790 align:start position:0% interesting part is also that it allows us<00:04:49.199> to<00:04:49.360> have<00:04:49.680> udfs<00:04:50.680> which<00:04:51.280> uh<00:04:51.479> actually 00:04:51.790 --> 00:04:51.800 align:start position:0% us to have udfs which uh actually 00:04:51.800 --> 00:04:53.830 align:start position:0% us to have udfs which uh actually encapsulate<00:04:52.479> a<00:04:52.560> lot<00:04:52.720> of<00:04:52.800> the<00:04:52.960> complex<00:04:53.360> logic 00:04:53.830 --> 00:04:53.840 align:start position:0% encapsulate a lot of the complex logic 00:04:53.840 --> 00:04:55.510 align:start position:0% encapsulate a lot of the complex logic that<00:04:54.160> you<00:04:54.280> may<00:04:54.440> not<00:04:54.560> be<00:04:54.720> able<00:04:54.880> to<00:04:55.120> write<00:04:55.320> in 00:04:55.510 --> 00:04:55.520 align:start position:0% that you may not be able to write in 00:04:55.520 --> 00:04:57.830 align:start position:0% that you may not be able to write in Hive<00:04:55.880> but<00:04:56.000> you<00:04:56.120> end<00:04:56.320> up<00:04:57.000> uh<00:04:57.440> like<00:04:57.560> let's<00:04:57.720> say 00:04:57.830 --> 00:04:57.840 align:start position:0% Hive but you end up uh like let's say 00:04:57.840 --> 00:04:59.990 align:start position:0% Hive but you end up uh like let's say you<00:04:57.919> wanted<00:04:58.160> to<00:04:58.320> apply<00:04:58.600> a<00:04:58.759> predictive<00:04:59.160> model 00:04:59.990 --> 00:05:00.000 align:start position:0% you wanted to apply a predictive model 00:05:00.000 --> 00:05:02.550 align:start position:0% you wanted to apply a predictive model uh<00:05:00.440> to<00:05:00.720> a<00:05:01.000> a<00:05:01.120> row<00:05:01.360> in<00:05:01.520> Hive<00:05:01.800> you<00:05:01.919> would<00:05:02.080> want<00:05:02.240> to 00:05:02.550 --> 00:05:02.560 align:start position:0% uh to a a row in Hive you would want to 00:05:02.560 --> 00:05:04.710 align:start position:0% uh to a a row in Hive you would want to have<00:05:02.720> that<00:05:02.840> as<00:05:02.960> a<00:05:03.039> UDF<00:05:03.600> that<00:05:03.759> can<00:05:04.360> maybe<00:05:04.600> look 00:05:04.710 --> 00:05:04.720 align:start position:0% have that as a UDF that can maybe look 00:05:04.720 --> 00:05:06.550 align:start position:0% have that as a UDF that can maybe look up<00:05:04.880> some<00:05:05.080> data<00:05:05.320> from<00:05:05.520> different<00:05:05.800> sources<00:05:06.360> do 00:05:06.550 --> 00:05:06.560 align:start position:0% up some data from different sources do 00:05:06.560 --> 00:05:08.909 align:start position:0% up some data from different sources do some<00:05:06.759> extraction<00:05:07.720> do<00:05:07.840> some<00:05:08.000> NLP<00:05:08.639> and<00:05:08.759> then 00:05:08.909 --> 00:05:08.919 align:start position:0% some extraction do some NLP and then 00:05:08.919 --> 00:05:10.950 align:start position:0% some extraction do some NLP and then apply<00:05:09.199> a<00:05:09.320> model<00:05:09.919> and<00:05:10.039> all<00:05:10.160> of<00:05:10.320> that<00:05:10.479> can<00:05:10.680> exist 00:05:10.950 --> 00:05:10.960 align:start position:0% apply a model and all of that can exist 00:05:10.960 --> 00:05:14.150 align:start position:0% apply a model and all of that can exist in<00:05:11.479> uh<00:05:11.639> a<00:05:11.759> UDF<00:05:12.320> so<00:05:13.039> I<00:05:13.120> think<00:05:13.280> that<00:05:13.600> combination 00:05:14.150 --> 00:05:14.160 align:start position:0% in uh a UDF so I think that combination 00:05:14.160 --> 00:05:16.350 align:start position:0% in uh a UDF so I think that combination of<00:05:14.479> the<00:05:14.639> Hadoop<00:05:15.000> stack<00:05:15.320> along<00:05:15.720> with<00:05:16.240> the 00:05:16.350 --> 00:05:16.360 align:start position:0% of the Hadoop stack along with the 00:05:16.360 --> 00:05:18.469 align:start position:0% of the Hadoop stack along with the ability<00:05:16.720> to<00:05:16.880> write<00:05:17.120> udfs<00:05:17.800> and<00:05:17.960> then<00:05:18.280> uh 00:05:18.469 --> 00:05:18.479 align:start position:0% ability to write udfs and then uh 00:05:18.479 --> 00:05:20.909 align:start position:0% ability to write udfs and then uh training<00:05:18.880> offline<00:05:19.240> models<00:05:19.639> using<00:05:20.520> uh<00:05:20.639> machine 00:05:20.909 --> 00:05:20.919 align:start position:0% training offline models using uh machine 00:05:20.919 --> 00:05:22.629 align:start position:0% training offline models using uh machine learning<00:05:21.199> tools<00:05:21.520> has<00:05:21.680> really<00:05:22.039> kind<00:05:22.160> of<00:05:22.360> helped 00:05:22.629 --> 00:05:22.639 align:start position:0% learning tools has really kind of helped 00:05:22.639 --> 00:05:25.870 align:start position:0% learning tools has really kind of helped us<00:05:22.800> in<00:05:23.319> uh<00:05:23.440> Skilling<00:05:23.919> up<00:05:24.680> cool<00:05:25.120> so<00:05:25.440> you<00:05:25.520> know<00:05:25.680> so 00:05:25.870 --> 00:05:25.880 align:start position:0% us in uh Skilling up cool so you know so 00:05:25.880 --> 00:05:27.110 align:start position:0% us in uh Skilling up cool so you know so when<00:05:26.000> you're<00:05:26.199> doing<00:05:26.479> it's<00:05:26.639> very<00:05:26.800> interesting 00:05:27.110 --> 00:05:27.120 align:start position:0% when you're doing it's very interesting 00:05:27.120 --> 00:05:28.510 align:start position:0% when you're doing it's very interesting because<00:05:27.240> you're<00:05:27.400> doing<00:05:27.639> you<00:05:27.720> know<00:05:27.880> using<00:05:28.240> both 00:05:28.510 --> 00:05:28.520 align:start position:0% because you're doing you know using both 00:05:28.520 --> 00:05:30.909 align:start position:0% because you're doing you know using both Python<00:05:28.800> and<00:05:28.960> weekend<00:05:29.880> uh<00:05:30.000> you<00:05:30.120> know<00:05:30.479> one<00:05:30.639> of<00:05:30.759> my 00:05:30.909 --> 00:05:30.919 align:start position:0% Python and weekend uh you know one of my 00:05:30.919 --> 00:05:32.550 align:start position:0% Python and weekend uh you know one of my huge<00:05:31.199> drives<00:05:31.600> in<00:05:31.720> the<00:05:31.800> scholar<00:05:32.160> Community<00:05:32.479> is 00:05:32.550 --> 00:05:32.560 align:start position:0% huge drives in the scholar Community is 00:05:32.560 --> 00:05:34.110 align:start position:0% huge drives in the scholar Community is to<00:05:32.680> kind<00:05:32.800> of<00:05:32.919> bring<00:05:33.240> scientific<00:05:33.639> comping<00:05:34.000> to 00:05:34.110 --> 00:05:34.120 align:start position:0% to kind of bring scientific comping to 00:05:34.120 --> 00:05:35.990 align:start position:0% to kind of bring scientific comping to the<00:05:34.240> jvm<00:05:34.960> right<00:05:35.199> right<00:05:35.360> and<00:05:35.479> so<00:05:35.680> there<00:05:35.800> is<00:05:35.919> a 00:05:35.990 --> 00:05:36.000 align:start position:0% the jvm right right and so there is a 00:05:36.000 --> 00:05:37.510 align:start position:0% the jvm right right and so there is a lot<00:05:36.120> of<00:05:36.240> libraries<00:05:36.639> which<00:05:36.800> exists<00:05:37.120> already 00:05:37.510 --> 00:05:37.520 align:start position:0% lot of libraries which exists already 00:05:37.520 --> 00:05:39.070 align:start position:0% lot of libraries which exists already but<00:05:37.800> there<00:05:37.919> are<00:05:38.039> some<00:05:38.240> gaps<00:05:38.520> compared<00:05:38.880> to 00:05:39.070 --> 00:05:39.080 align:start position:0% but there are some gaps compared to 00:05:39.080 --> 00:05:41.469 align:start position:0% but there are some gaps compared to Python<00:05:39.400> and<00:05:39.840> there<00:05:39.960> is<00:05:40.120> just<00:05:40.240> a<00:05:40.479> habit<00:05:40.840> of<00:05:41.400> a 00:05:41.469 --> 00:05:41.479 align:start position:0% Python and there is just a habit of a 00:05:41.479 --> 00:05:43.150 align:start position:0% Python and there is just a habit of a lot<00:05:41.680> of<00:05:41.840> data<00:05:42.080> scientists<00:05:42.440> to<00:05:42.560> use<00:05:42.800> Python 00:05:43.150 --> 00:05:43.160 align:start position:0% lot of data scientists to use Python 00:05:43.160 --> 00:05:44.350 align:start position:0% lot of data scientists to use Python because<00:05:43.360> that's<00:05:43.720> usually<00:05:44.000> the<00:05:44.120> first 00:05:44.350 --> 00:05:44.360 align:start position:0% because that's usually the first 00:05:44.360 --> 00:05:46.430 align:start position:0% because that's usually the first language<00:05:45.240> uh<00:05:45.360> they<00:05:45.479> pick<00:05:45.680> up<00:05:45.880> in<00:05:46.039> in<00:05:46.120> school 00:05:46.430 --> 00:05:46.440 align:start position:0% language uh they pick up in in school 00:05:46.440 --> 00:05:47.990 align:start position:0% language uh they pick up in in school it's<00:05:46.560> also<00:05:46.759> probably<00:05:47.000> the<00:05:47.120> most<00:05:47.400> developed<00:05:47.880> in 00:05:47.990 --> 00:05:48.000 align:start position:0% it's also probably the most developed in 00:05:48.000 --> 00:05:50.550 align:start position:0% it's also probably the most developed in terms<00:05:48.280> of<00:05:48.600> machine<00:05:48.880> learning<00:05:49.319> in<00:05:49.960> uh<00:05:50.199> because 00:05:50.550 --> 00:05:50.560 align:start position:0% terms of machine learning in uh because 00:05:50.560 --> 00:05:52.270 align:start position:0% terms of machine learning in uh because there<00:05:50.759> there<00:05:50.840> are<00:05:50.960> so<00:05:51.080> many<00:05:51.319> tools<00:05:51.800> that<00:05:51.919> are 00:05:52.270 --> 00:05:52.280 align:start position:0% there there are so many tools that are 00:05:52.280 --> 00:05:54.990 align:start position:0% there there are so many tools that are present<00:05:52.600> over<00:05:52.840> there<00:05:53.080> but<00:05:53.479> yeah<00:05:53.919> right<00:05:54.360> so<00:05:54.639> and 00:05:54.990 --> 00:05:55.000 align:start position:0% present over there but yeah right so and 00:05:55.000 --> 00:05:56.629 align:start position:0% present over there but yeah right so and so<00:05:55.240> that<00:05:55.440> you<00:05:55.560> know<00:05:55.720> that's<00:05:55.919> my<00:05:56.080> question<00:05:56.400> to 00:05:56.629 --> 00:05:56.639 align:start position:0% so that you know that's my question to 00:05:56.639 --> 00:05:59.309 align:start position:0% so that you know that's my question to you<00:05:56.759> is<00:05:56.919> kind<00:05:57.039> of<00:05:57.240> practicing<00:05:58.120> with<00:05:58.360> both<00:05:59.039> uh 00:05:59.309 --> 00:05:59.319 align:start position:0% you is kind of practicing with both uh 00:05:59.319 --> 00:06:03.150 align:start position:0% you is kind of practicing with both uh to<00:05:59.639> tool<00:05:59.840> sets<00:06:00.560> what<00:06:00.840> does<00:06:01.280> the<00:06:02.120> jvm<00:06:02.720> community 00:06:03.150 --> 00:06:03.160 align:start position:0% to tool sets what does the jvm community 00:06:03.160 --> 00:06:07.430 align:start position:0% to tool sets what does the jvm community need<00:06:03.400> to<00:06:03.639> do<00:06:04.520> to<00:06:05.120> kind<00:06:05.280> of<00:06:05.440> bring<00:06:05.759> the<00:06:06.000> python 00:06:07.430 --> 00:06:07.440 align:start position:0% need to do to kind of bring the python 00:06:07.440 --> 00:06:09.670 align:start position:0% need to do to kind of bring the python functionality<00:06:08.440> right<00:06:08.680> where<00:06:08.800> do<00:06:08.919> you<00:06:09.000> see<00:06:09.280> the 00:06:09.670 --> 00:06:09.680 align:start position:0% functionality right where do you see the 00:06:09.680 --> 00:06:12.710 align:start position:0% functionality right where do you see the most<00:06:10.680> gaps<00:06:11.120> and<00:06:11.440> where<00:06:11.759> the<00:06:11.919> community<00:06:12.319> can 00:06:12.710 --> 00:06:12.720 align:start position:0% most gaps and where the community can 00:06:12.720 --> 00:06:14.909 align:start position:0% most gaps and where the community can contribute<00:06:13.199> the<00:06:13.360> most<00:06:13.720> and<00:06:13.919> you<00:06:14.039> know<00:06:14.560> what 00:06:14.909 --> 00:06:14.919 align:start position:0% contribute the most and you know what 00:06:14.919 --> 00:06:16.870 align:start position:0% contribute the most and you know what should<00:06:15.199> happen<00:06:15.599> so<00:06:15.800> for<00:06:15.960> you<00:06:16.080> to<00:06:16.319> stay<00:06:16.639> within 00:06:16.870 --> 00:06:16.880 align:start position:0% should happen so for you to stay within 00:06:16.880 --> 00:06:18.950 align:start position:0% should happen so for you to stay within the<00:06:17.000> jvm<00:06:17.759> when<00:06:17.960> you<00:06:18.199> you<00:06:18.319> know<00:06:18.520> do<00:06:18.720> your 00:06:18.950 --> 00:06:18.960 align:start position:0% the jvm when you you know do your 00:06:18.960 --> 00:06:21.870 align:start position:0% the jvm when you you know do your experiments<00:06:19.919> right<00:06:20.440> uh<00:06:20.840> I<00:06:20.919> think<00:06:21.120> the<00:06:21.280> main 00:06:21.870 --> 00:06:21.880 align:start position:0% experiments right uh I think the main 00:06:21.880 --> 00:06:24.350 align:start position:0% experiments right uh I think the main thing<00:06:22.120> is<00:06:22.280> the<00:06:22.479> comprehensive<00:06:23.319> nature<00:06:23.880> of<00:06:24.199> the 00:06:24.350 --> 00:06:24.360 align:start position:0% thing is the comprehensive nature of the 00:06:24.360 --> 00:06:26.790 align:start position:0% thing is the comprehensive nature of the tools<00:06:24.680> that<00:06:24.800> are<00:06:25.039> present<00:06:25.840> in<00:06:26.199> Python<00:06:26.560> I<00:06:26.639> think 00:06:26.790 --> 00:06:26.800 align:start position:0% tools that are present in Python I think 00:06:26.800 --> 00:06:28.390 align:start position:0% tools that are present in Python I think a<00:06:26.919> lot<00:06:27.039> of<00:06:27.160> the<00:06:27.319> algorithms<00:06:27.840> that<00:06:27.960> are<00:06:28.120> present 00:06:28.390 --> 00:06:28.400 align:start position:0% a lot of the algorithms that are present 00:06:28.400 --> 00:06:30.510 align:start position:0% a lot of the algorithms that are present in<00:06:28.560> like<00:06:28.680> the<00:06:28.840> python<00:06:29.080> toolkits<00:06:29.919> are<00:06:30.120> not 00:06:30.510 --> 00:06:30.520 align:start position:0% in like the python toolkits are not 00:06:30.520 --> 00:06:34.670 align:start position:0% in like the python toolkits are not always<00:06:31.440> in<00:06:32.319> the<00:06:32.479> jvm<00:06:32.919> toolkits<00:06:33.440> and<00:06:33.960> maybe<00:06:34.520> if 00:06:34.670 --> 00:06:34.680 align:start position:0% always in the jvm toolkits and maybe if 00:06:34.680 --> 00:06:36.189 align:start position:0% always in the jvm toolkits and maybe if they<00:06:34.800> are<00:06:35.039> present<00:06:35.319> they're<00:06:35.520> not<00:06:35.680> implemented 00:06:36.189 --> 00:06:36.199 align:start position:0% they are present they're not implemented 00:06:36.199 --> 00:06:38.430 align:start position:0% they are present they're not implemented in<00:06:36.280> a<00:06:36.440> fashion<00:06:36.800> that<00:06:36.919> is<00:06:37.120> very<00:06:37.759> uh<00:06:37.919> efficient 00:06:38.430 --> 00:06:38.440 align:start position:0% in a fashion that is very uh efficient 00:06:38.440 --> 00:06:40.950 align:start position:0% in a fashion that is very uh efficient when<00:06:38.599> you<00:06:38.720> want<00:06:38.880> to<00:06:39.000> scale<00:06:39.319> up<00:06:39.960> so<00:06:40.479> uh<00:06:40.639> I<00:06:40.720> think 00:06:40.950 --> 00:06:40.960 align:start position:0% when you want to scale up so uh I think 00:06:40.960 --> 00:06:44.670 align:start position:0% when you want to scale up so uh I think a<00:06:41.120> lot<00:06:41.280> of<00:06:41.800> uh<00:06:42.520> uh<00:06:43.440> there<00:06:43.560> are<00:06:43.680> a<00:06:43.800> lot<00:06:43.960> of<00:06:44.160> small 00:06:44.670 --> 00:06:44.680 align:start position:0% a lot of uh uh there are a lot of small 00:06:44.680 --> 00:06:47.510 align:start position:0% a lot of uh uh there are a lot of small things<00:06:45.479> which<00:06:46.440> the<00:06:46.560> smaller<00:06:46.919> pieces<00:06:47.280> once 00:06:47.510 --> 00:06:47.520 align:start position:0% things which the smaller pieces once 00:06:47.520 --> 00:06:49.029 align:start position:0% things which the smaller pieces once they<00:06:47.639> are<00:06:47.840> there<00:06:48.039> and<00:06:48.199> they<00:06:48.400> kind<00:06:48.520> of<00:06:48.680> build<00:06:48.880> up 00:06:49.029 --> 00:06:49.039 align:start position:0% they are there and they kind of build up 00:06:49.039 --> 00:06:51.230 align:start position:0% they are there and they kind of build up like<00:06:49.199> Lego<00:06:49.440> bricks<00:06:50.199> that's<00:06:50.360> when<00:06:50.919> maybe<00:06:51.160> the 00:06:51.230 --> 00:06:51.240 align:start position:0% like Lego bricks that's when maybe the 00:06:51.240 --> 00:06:53.950 align:start position:0% like Lego bricks that's when maybe the jvm<00:06:52.039> uh<00:06:52.199> Community<00:06:52.639> would<00:06:52.840> be<00:06:53.400> at<00:06:53.639> that<00:06:53.800> same 00:06:53.950 --> 00:06:53.960 align:start position:0% jvm uh Community would be at that same 00:06:53.960 --> 00:06:56.469 align:start position:0% jvm uh Community would be at that same scale<00:06:54.680> uh<00:06:54.840> I'm<00:06:55.039> looking<00:06:55.280> forward<00:06:55.599> to<00:06:56.120> using 00:06:56.469 --> 00:06:56.479 align:start position:0% scale uh I'm looking forward to using 00:06:56.479 --> 00:06:58.550 align:start position:0% scale uh I'm looking forward to using spark<00:06:56.879> MLB<00:06:57.440> I<00:06:57.560> haven't<00:06:57.759> used<00:06:57.960> it<00:06:58.120> personally 00:06:58.550 --> 00:06:58.560 align:start position:0% spark MLB I haven't used it personally 00:06:58.560 --> 00:07:00.749 align:start position:0% spark MLB I haven't used it personally yet<00:06:58.800> but<00:06:59.199> that's<00:06:59.560> seems<00:06:59.720> to<00:06:59.800> be<00:06:59.879> a<00:07:00.000> promising 00:07:00.749 --> 00:07:00.759 align:start position:0% yet but that's seems to be a promising 00:07:00.759 --> 00:07:02.550 align:start position:0% yet but that's seems to be a promising area<00:07:01.160> which<00:07:01.400> could<00:07:01.639> combine<00:07:02.120> like<00:07:02.360> the 00:07:02.550 --> 00:07:02.560 align:start position:0% area which could combine like the 00:07:02.560 --> 00:07:05.029 align:start position:0% area which could combine like the advantages<00:07:03.000> of<00:07:03.120> streaming<00:07:03.560> data<00:07:03.840> along<00:07:04.160> with 00:07:05.029 --> 00:07:05.039 align:start position:0% advantages of streaming data along with 00:07:05.039 --> 00:07:06.110 align:start position:0% advantages of streaming data along with uh<00:07:05.240> some<00:07:05.400> of<00:07:05.479> the<00:07:05.599> machine<00:07:05.840> learning 00:07:06.110 --> 00:07:06.120 align:start position:0% uh some of the machine learning 00:07:06.120 --> 00:07:08.390 align:start position:0% uh some of the machine learning libraries<00:07:06.639> so<00:07:07.360> uh<00:07:07.680> I'm<00:07:07.759> sure<00:07:08.000> like<00:07:08.160> there<00:07:08.240> are 00:07:08.390 --> 00:07:08.400 align:start position:0% libraries so uh I'm sure like there are 00:07:08.400 --> 00:07:12.430 align:start position:0% libraries so uh I'm sure like there are a<00:07:08.479> lot<00:07:08.639> of<00:07:08.919> avenues<00:07:09.560> for<00:07:10.560> uh<00:07:10.919> implementing<00:07:11.759> a<00:07:12.360> a 00:07:12.430 --> 00:07:12.440 align:start position:0% a lot of avenues for uh implementing a a 00:07:12.440 --> 00:07:13.950 align:start position:0% a lot of avenues for uh implementing a a lot<00:07:12.599> of<00:07:12.720> the<00:07:12.840> things<00:07:13.000> that<00:07:13.160> are<00:07:13.520> existing<00:07:13.840> in 00:07:13.950 --> 00:07:13.960 align:start position:0% lot of the things that are existing in 00:07:13.960 --> 00:07:16.629 align:start position:0% lot of the things that are existing in the<00:07:14.120> python<00:07:14.440> toolkits<00:07:15.120> in<00:07:15.360> into<00:07:15.800> the<00:07:16.160> jvm 00:07:16.629 --> 00:07:16.639 align:start position:0% the python toolkits in into the jvm 00:07:16.639 --> 00:07:18.710 align:start position:0% the python toolkits in into the jvm space<00:07:17.080> yeah<00:07:17.199> and<00:07:17.360> Spark<00:07:17.720> also<00:07:17.919> will<00:07:18.080> run<00:07:18.319> Hive 00:07:18.710 --> 00:07:18.720 align:start position:0% space yeah and Spark also will run Hive 00:07:18.720 --> 00:07:20.510 align:start position:0% space yeah and Spark also will run Hive right<00:07:18.879> because<00:07:19.120> you<00:07:19.240> know<00:07:19.919> are<00:07:20.039> committed<00:07:20.400> to 00:07:20.510 --> 00:07:20.520 align:start position:0% right because you know are committed to 00:07:20.520 --> 00:07:22.990 align:start position:0% right because you know are committed to run<00:07:20.960> everything<00:07:21.560> on<00:07:21.759> spark<00:07:22.560> so<00:07:22.759> you<00:07:22.879> know 00:07:22.990 --> 00:07:23.000 align:start position:0% run everything on spark so you know 00:07:23.000 --> 00:07:25.150 align:start position:0% run everything on spark so you know there<00:07:23.080> is<00:07:23.240> spark<00:07:23.479> cql<00:07:23.879> but<00:07:24.000> also<00:07:24.479> Hive<00:07:24.800> itself 00:07:25.150 --> 00:07:25.160 align:start position:0% there is spark cql but also Hive itself 00:07:25.160 --> 00:07:27.110 align:start position:0% there is spark cql but also Hive itself will<00:07:25.319> will<00:07:25.520> have<00:07:25.720> spark<00:07:26.479> back<00:07:26.639> end<00:07:26.879> right<00:07:27.000> so 00:07:27.110 --> 00:07:27.120 align:start position:0% will will have spark back end right so 00:07:27.120 --> 00:07:28.830 align:start position:0% will will have spark back end right so which<00:07:27.240> is<00:07:27.400> which<00:07:27.479> is<00:07:27.599> really<00:07:27.840> exciting<00:07:28.520> yes 00:07:28.830 --> 00:07:28.840 align:start position:0% which is which is really exciting yes 00:07:28.840 --> 00:07:31.270 align:start position:0% which is which is really exciting yes yeah<00:07:29.319> um<00:07:29.599> interesting<00:07:30.000> interesting<00:07:30.680> so<00:07:31.080> uh<00:07:31.160> so 00:07:31.270 --> 00:07:31.280 align:start position:0% yeah um interesting interesting so uh so 00:07:31.280 --> 00:07:32.710 align:start position:0% yeah um interesting interesting so uh so you<00:07:31.360> mentioned<00:07:31.680> NLP<00:07:32.039> can<00:07:32.160> you<00:07:32.360> talk<00:07:32.479> a<00:07:32.560> little 00:07:32.710 --> 00:07:32.720 align:start position:0% you mentioned NLP can you talk a little 00:07:32.720 --> 00:07:35.390 align:start position:0% you mentioned NLP can you talk a little bit<00:07:32.960> about<00:07:33.319> what<00:07:33.479> kind<00:07:33.599> of<00:07:33.720> NLP<00:07:34.560> uh<00:07:35.120> are<00:07:35.240> you 00:07:35.390 --> 00:07:35.400 align:start position:0% bit about what kind of NLP uh are you 00:07:35.400 --> 00:07:36.749 align:start position:0% bit about what kind of NLP uh are you guys<00:07:35.560> doing<00:07:35.919> what<00:07:36.039> kind<00:07:36.160> of<00:07:36.240> tools<00:07:36.520> are<00:07:36.639> you 00:07:36.749 --> 00:07:36.759 align:start position:0% guys doing what kind of tools are you 00:07:36.759 --> 00:07:39.990 align:start position:0% guys doing what kind of tools are you using<00:07:37.319> and<00:07:37.639> what<00:07:37.879> works<00:07:38.199> in<00:07:38.400> practice<00:07:39.240> Yeah<00:07:39.520> so 00:07:39.990 --> 00:07:40.000 align:start position:0% using and what works in practice Yeah so 00:07:40.000 --> 00:07:41.830 align:start position:0% using and what works in practice Yeah so uh<00:07:40.160> I<00:07:40.280> think<00:07:40.440> one<00:07:40.560> of<00:07:40.680> the<00:07:40.840> biggest<00:07:41.479> uh<00:07:41.599> things 00:07:41.830 --> 00:07:41.840 align:start position:0% uh I think one of the biggest uh things 00:07:41.840 --> 00:07:44.589 align:start position:0% uh I think one of the biggest uh things that<00:07:42.000> we<00:07:42.240> found<00:07:42.560> in<00:07:42.720> our<00:07:43.400> uh<00:07:43.919> experiments<00:07:44.440> is 00:07:44.589 --> 00:07:44.599 align:start position:0% that we found in our uh experiments is 00:07:44.599 --> 00:07:47.309 align:start position:0% that we found in our uh experiments is that<00:07:45.599> if<00:07:45.720> you<00:07:45.879> try<00:07:46.120> to<00:07:46.280> use<00:07:46.560> like<00:07:46.800> s<00:07:47.120> very 00:07:47.309 --> 00:07:47.319 align:start position:0% that if you try to use like s very 00:07:47.319 --> 00:07:49.430 align:start position:0% that if you try to use like s very sophisticated<00:07:48.120> NLP<00:07:48.639> techniques<00:07:49.120> like<00:07:49.280> let's 00:07:49.430 --> 00:07:49.440 align:start position:0% sophisticated NLP techniques like let's 00:07:49.440 --> 00:07:50.830 align:start position:0% sophisticated NLP techniques like let's say<00:07:49.560> you<00:07:49.639> wanted<00:07:49.879> to<00:07:49.960> use<00:07:50.159> Stanford<00:07:50.560> named 00:07:50.830 --> 00:07:50.840 align:start position:0% say you wanted to use Stanford named 00:07:50.840 --> 00:07:53.670 align:start position:0% say you wanted to use Stanford named entity<00:07:51.159> recognition<00:07:52.159> for<00:07:52.720> passing<00:07:53.080> out<00:07:53.479> uh 00:07:53.670 --> 00:07:53.680 align:start position:0% entity recognition for passing out uh 00:07:53.680 --> 00:07:56.589 align:start position:0% entity recognition for passing out uh entities<00:07:54.120> from<00:07:54.280> your<00:07:54.479> text<00:07:55.440> uh<00:07:55.720> the<00:07:56.440> what 00:07:56.589 --> 00:07:56.599 align:start position:0% entities from your text uh the what 00:07:56.599 --> 00:07:58.390 align:start position:0% entities from your text uh the what happens<00:07:56.960> is<00:07:57.159> that<00:07:57.520> sure<00:07:57.800> you<00:07:57.879> could<00:07:58.039> run<00:07:58.280> that 00:07:58.390 --> 00:07:58.400 align:start position:0% happens is that sure you could run that 00:07:58.400 --> 00:08:00.110 align:start position:0% happens is that sure you could run that on<00:07:58.520> a<00:07:58.680> small<00:07:58.960> scale<00:07:59.280> but<00:07:59.560> you<00:07:59.840> when<00:07:59.919> you<00:08:00.039> want 00:08:00.110 --> 00:08:00.120 align:start position:0% on a small scale but you when you want 00:08:00.120 --> 00:08:03.230 align:start position:0% on a small scale but you when you want to<00:08:00.240> run<00:08:00.479> something<00:08:00.879> on<00:08:01.840> 700<00:08:02.440> million<00:08:02.800> messages 00:08:03.230 --> 00:08:03.240 align:start position:0% to run something on 700 million messages 00:08:03.240 --> 00:08:04.950 align:start position:0% to run something on 700 million messages per<00:08:03.440> day<00:08:03.960> it's<00:08:04.120> not<00:08:04.240> going<00:08:04.360> to<00:08:04.440> scale<00:08:04.759> very 00:08:04.950 --> 00:08:04.960 align:start position:0% per day it's not going to scale very 00:08:04.960 --> 00:08:07.869 align:start position:0% per day it's not going to scale very well<00:08:05.800> uh<00:08:06.039> so<00:08:06.680> uh<00:08:06.879> one<00:08:07.000> of<00:08:07.159> the<00:08:07.280> approaches<00:08:07.720> that 00:08:07.869 --> 00:08:07.879 align:start position:0% well uh so uh one of the approaches that 00:08:07.879 --> 00:08:11.070 align:start position:0% well uh so uh one of the approaches that we<00:08:08.080> took<00:08:08.440> was<00:08:08.720> that<00:08:09.120> we<00:08:09.520> actually<00:08:10.199> used<00:08:10.879> uh 00:08:11.070 --> 00:08:11.080 align:start position:0% we took was that we actually used uh 00:08:11.080 --> 00:08:13.430 align:start position:0% we took was that we actually used uh open<00:08:11.440> data<00:08:11.759> sets<00:08:12.120> such<00:08:12.319> as<00:08:12.680> uh<00:08:12.800> Freebase 00:08:13.430 --> 00:08:13.440 align:start position:0% open data sets such as uh Freebase 00:08:13.440 --> 00:08:15.469 align:start position:0% open data sets such as uh Freebase Wikipedia<00:08:14.440> and<00:08:14.560> all<00:08:14.680> of<00:08:14.840> these<00:08:15.000> knowledge 00:08:15.469 --> 00:08:15.479 align:start position:0% Wikipedia and all of these knowledge 00:08:15.479 --> 00:08:17.950 align:start position:0% Wikipedia and all of these knowledge grafts<00:08:16.400> and<00:08:16.680> using<00:08:17.039> those<00:08:17.240> knowledge<00:08:17.599> crafts 00:08:17.950 --> 00:08:17.960 align:start position:0% grafts and using those knowledge crafts 00:08:17.960 --> 00:08:20.589 align:start position:0% grafts and using those knowledge crafts we<00:08:18.120> try<00:08:18.319> to<00:08:18.440> create<00:08:18.840> dictionaries<00:08:19.840> that<00:08:20.080> are 00:08:20.589 --> 00:08:20.599 align:start position:0% we try to create dictionaries that are 00:08:20.599 --> 00:08:23.909 align:start position:0% we try to create dictionaries that are uh<00:08:21.199> and<00:08:21.479> that<00:08:21.560> are<00:08:21.759> pre-trained<00:08:22.319> and<00:08:22.919> pre 00:08:23.909 --> 00:08:23.919 align:start position:0% uh and that are pre-trained and pre 00:08:23.919 --> 00:08:26.230 align:start position:0% uh and that are pre-trained and pre pre-fabricated<00:08:24.680> so<00:08:24.840> that<00:08:25.000> they<00:08:25.120> are<00:08:26.120> they 00:08:26.230 --> 00:08:26.240 align:start position:0% pre-fabricated so that they are they 00:08:26.240 --> 00:08:28.029 align:start position:0% pre-fabricated so that they are they already<00:08:26.560> encapsulate<00:08:27.199> all<00:08:27.360> the<00:08:27.560> entities 00:08:28.029 --> 00:08:28.039 align:start position:0% already encapsulate all the entities 00:08:28.039 --> 00:08:30.510 align:start position:0% already encapsulate all the entities that<00:08:28.199> are<00:08:28.400> relevant<00:08:29.159> and<00:08:29.879> the<00:08:30.039> advantage<00:08:30.400> of 00:08:30.510 --> 00:08:30.520 align:start position:0% that are relevant and the advantage of 00:08:30.520 --> 00:08:32.149 align:start position:0% that are relevant and the advantage of doing<00:08:30.800> that<00:08:30.919> is<00:08:31.080> also<00:08:31.440> that<00:08:31.759> when<00:08:31.960> new 00:08:32.149 --> 00:08:32.159 align:start position:0% doing that is also that when new 00:08:32.159 --> 00:08:33.709 align:start position:0% doing that is also that when new entities<00:08:32.560> start<00:08:32.760> coming<00:08:32.959> up<00:08:33.360> let's<00:08:33.560> say 00:08:33.709 --> 00:08:33.719 align:start position:0% entities start coming up let's say 00:08:33.719 --> 00:08:35.550 align:start position:0% entities start coming up let's say there's<00:08:33.880> a<00:08:34.279> completely<00:08:34.719> new<00:08:34.880> phenomenon<00:08:35.320> on 00:08:35.550 --> 00:08:35.560 align:start position:0% there's a completely new phenomenon on 00:08:35.560 --> 00:08:37.790 align:start position:0% there's a completely new phenomenon on social<00:08:35.800> media<00:08:36.120> today<00:08:36.880> that<00:08:37.360> we<00:08:37.519> want<00:08:37.680> to 00:08:37.790 --> 00:08:37.800 align:start position:0% social media today that we want to 00:08:37.800 --> 00:08:39.149 align:start position:0% social media today that we want to incorporate<00:08:38.200> in<00:08:38.320> our<00:08:38.440> dictionaries<00:08:38.919> we<00:08:39.039> are 00:08:39.149 --> 00:08:39.159 align:start position:0% incorporate in our dictionaries we are 00:08:39.159 --> 00:08:40.509 align:start position:0% incorporate in our dictionaries we are able<00:08:39.320> to<00:08:39.440> update<00:08:39.680> our<00:08:39.839> dictionaries<00:08:40.360> very 00:08:40.509 --> 00:08:40.519 align:start position:0% able to update our dictionaries very 00:08:40.519 --> 00:08:44.110 align:start position:0% able to update our dictionaries very easily<00:08:41.360> so<00:08:41.919> rather<00:08:42.159> than<00:08:42.399> trying<00:08:42.719> to<00:08:43.120> extract 00:08:44.110 --> 00:08:44.120 align:start position:0% easily so rather than trying to extract 00:08:44.120 --> 00:08:48.070 align:start position:0% easily so rather than trying to extract uh<00:08:44.560> semantic<00:08:45.320> meaning<00:08:46.320> in<00:08:46.680> uh<00:08:47.640> of<00:08:47.760> course 00:08:48.070 --> 00:08:48.080 align:start position:0% uh semantic meaning in uh of course 00:08:48.080 --> 00:08:50.030 align:start position:0% uh semantic meaning in uh of course sometimes<00:08:48.320> that<00:08:48.440> is<00:08:48.600> very<00:08:48.880> useful<00:08:49.399> but<00:08:49.800> in 00:08:50.030 --> 00:08:50.040 align:start position:0% sometimes that is very useful but in 00:08:50.040 --> 00:08:52.470 align:start position:0% sometimes that is very useful but in some<00:08:50.320> cases<00:08:51.120> what<00:08:51.399> happens<00:08:51.720> is<00:08:51.920> just<00:08:52.040> using<00:08:52.360> a 00:08:52.470 --> 00:08:52.480 align:start position:0% some cases what happens is just using a 00:08:52.480 --> 00:08:54.110 align:start position:0% some cases what happens is just using a simple<00:08:52.800> dictionary<00:08:53.320> and<00:08:53.560> trying<00:08:53.760> to<00:08:53.920> have 00:08:54.110 --> 00:08:54.120 align:start position:0% simple dictionary and trying to have 00:08:54.120 --> 00:08:55.470 align:start position:0% simple dictionary and trying to have those<00:08:54.320> dictionary<00:08:54.760> terms<00:08:55.040> mapping<00:08:55.320> to 00:08:55.470 --> 00:08:55.480 align:start position:0% those dictionary terms mapping to 00:08:55.480 --> 00:08:56.910 align:start position:0% those dictionary terms mapping to certain<00:08:55.720> entities<00:08:56.080> in<00:08:56.240> your<00:08:56.399> topical 00:08:56.910 --> 00:08:56.920 align:start position:0% certain entities in your topical 00:08:56.920 --> 00:09:00.910 align:start position:0% certain entities in your topical ontology<00:08:57.920> are<00:08:58.519> uh<00:08:59.200> is<00:08:59.480> pretty<00:08:59.760> effective<00:09:00.320> in 00:09:00.910 --> 00:09:00.920 align:start position:0% ontology are uh is pretty effective in 00:09:00.920 --> 00:09:04.069 align:start position:0% ontology are uh is pretty effective in uh<00:09:01.399> going<00:09:01.880> in<00:09:02.240> going<00:09:02.480> maybe<00:09:02.720> 90%<00:09:03.240> of<00:09:03.360> the<00:09:03.480> way 00:09:04.069 --> 00:09:04.079 align:start position:0% uh going in going maybe 90% of the way 00:09:04.079 --> 00:09:05.509 align:start position:0% uh going in going maybe 90% of the way and<00:09:04.240> sure<00:09:04.519> at<00:09:04.600> the<00:09:04.720> end<00:09:04.839> of<00:09:04.959> the<00:09:05.079> day<00:09:05.240> if<00:09:05.320> you 00:09:05.509 --> 00:09:05.519 align:start position:0% and sure at the end of the day if you 00:09:05.519 --> 00:09:07.150 align:start position:0% and sure at the end of the day if you really<00:09:05.720> want<00:09:05.920> to<00:09:06.040> be<00:09:06.279> like<00:09:06.640> very<00:09:06.880> highly 00:09:07.150 --> 00:09:07.160 align:start position:0% really want to be like very highly 00:09:07.160 --> 00:09:09.590 align:start position:0% really want to be like very highly accurate<00:09:07.560> you<00:09:07.720> have<00:09:07.839> to<00:09:08.640> uh<00:09:09.200> get<00:09:09.360> more 00:09:09.590 --> 00:09:09.600 align:start position:0% accurate you have to uh get more 00:09:09.600 --> 00:09:11.470 align:start position:0% accurate you have to uh get more sophisticated<00:09:10.279> in<00:09:10.440> terms<00:09:10.680> of<00:09:10.880> the<00:09:11.040> NLP 00:09:11.470 --> 00:09:11.480 align:start position:0% sophisticated in terms of the NLP 00:09:11.480 --> 00:09:13.590 align:start position:0% sophisticated in terms of the NLP techniques<00:09:11.959> but<00:09:12.440> even<00:09:12.880> doing<00:09:13.240> Simple 00:09:13.590 --> 00:09:13.600 align:start position:0% techniques but even doing Simple 00:09:13.600 --> 00:09:15.829 align:start position:0% techniques but even doing Simple techniques<00:09:14.000> at<00:09:14.200> scale<00:09:15.000> uh<00:09:15.120> can<00:09:15.519> sometimes<00:09:15.720> get 00:09:15.829 --> 00:09:15.839 align:start position:0% techniques at scale uh can sometimes get 00:09:15.839 --> 00:09:17.110 align:start position:0% techniques at scale uh can sometimes get you<00:09:15.959> a<00:09:16.079> long<00:09:16.279> way<00:09:16.480> if<00:09:16.560> you're<00:09:16.720> using<00:09:17.000> the 00:09:17.110 --> 00:09:17.120 align:start position:0% you a long way if you're using the 00:09:17.120 --> 00:09:18.710 align:start position:0% you a long way if you're using the correct<00:09:17.399> kind<00:09:17.560> of<00:09:17.720> data<00:09:17.959> sets<00:09:18.240> and<00:09:18.360> the<00:09:18.519> data 00:09:18.710 --> 00:09:18.720 align:start position:0% correct kind of data sets and the data 00:09:18.720 --> 00:09:21.069 align:start position:0% correct kind of data sets and the data sources<00:09:19.440> interesting<00:09:20.079> so<00:09:20.399> so<00:09:20.600> basically<00:09:20.920> you 00:09:21.069 --> 00:09:21.079 align:start position:0% sources interesting so so basically you 00:09:21.079 --> 00:09:23.870 align:start position:0% sources interesting so so basically you do<00:09:21.360> your<00:09:21.600> most<00:09:21.880> of<00:09:22.000> yourp<00:09:22.720> offline<00:09:23.480> and<00:09:23.680> and 00:09:23.870 --> 00:09:23.880 align:start position:0% do your most of yourp offline and and 00:09:23.880 --> 00:09:26.030 align:start position:0% do your most of yourp offline and and you<00:09:24.120> when<00:09:24.200> you<00:09:24.279> do<00:09:24.440> training<00:09:24.920> so<00:09:25.560> uh<00:09:25.720> when<00:09:25.880> you 00:09:26.030 --> 00:09:26.040 align:start position:0% you when you do training so uh when you 00:09:26.040 --> 00:09:28.389 align:start position:0% you when you do training so uh when you do<00:09:26.320> that<00:09:27.160> can<00:09:27.279> you<00:09:27.480> describe<00:09:27.920> what<00:09:28.120> kind<00:09:28.279> of 00:09:28.389 --> 00:09:28.399 align:start position:0% do that can you describe what kind of 00:09:28.399 --> 00:09:31.470 align:start position:0% do that can you describe what kind of Stack<00:09:28.720> are<00:09:28.800> you<00:09:28.920> using<00:09:29.720> sure<00:09:30.399> uh<00:09:30.600> so<00:09:30.959> basically 00:09:31.470 --> 00:09:31.480 align:start position:0% Stack are you using sure uh so basically 00:09:31.480 --> 00:09:34.190 align:start position:0% Stack are you using sure uh so basically we<00:09:31.680> use<00:09:32.160> uh<00:09:32.279> Freebase<00:09:32.880> as<00:09:33.040> one<00:09:33.160> of<00:09:33.279> our<00:09:33.600> primary 00:09:34.190 --> 00:09:34.200 align:start position:0% we use uh Freebase as one of our primary 00:09:34.200 --> 00:09:37.030 align:start position:0% we use uh Freebase as one of our primary uh<00:09:34.360> data<00:09:34.600> sources<00:09:35.399> so<00:09:35.959> again<00:09:36.160> we<00:09:36.320> use<00:09:36.600> Hive<00:09:36.880> to 00:09:37.030 --> 00:09:37.040 align:start position:0% uh data sources so again we use Hive to 00:09:37.040 --> 00:09:39.910 align:start position:0% uh data sources so again we use Hive to kind<00:09:37.160> of<00:09:37.440> extract<00:09:37.959> all<00:09:38.160> these<00:09:38.560> uh<00:09:38.800> entities<00:09:39.800> uh 00:09:39.910 --> 00:09:39.920 align:start position:0% kind of extract all these uh entities uh 00:09:39.920 --> 00:09:42.269 align:start position:0% kind of extract all these uh entities uh we<00:09:40.079> have<00:09:40.240> a<00:09:40.360> bunch<00:09:40.560> of<00:09:40.959> tools<00:09:41.399> that<00:09:41.519> we<00:09:41.640> have 00:09:42.269 --> 00:09:42.279 align:start position:0% we have a bunch of tools that we have 00:09:42.279 --> 00:09:44.829 align:start position:0% we have a bunch of tools that we have that<00:09:43.200> in<00:09:43.320> the<00:09:43.440> udfs<00:09:44.240> that<00:09:44.360> do<00:09:44.519> things<00:09:44.680> like 00:09:44.829 --> 00:09:44.839 align:start position:0% that in the udfs that do things like 00:09:44.839 --> 00:09:47.630 align:start position:0% that in the udfs that do things like tokenization<00:09:45.720> stemming<00:09:46.720> and<00:09:46.880> maybe<00:09:47.200> a<00:09:47.320> lot<00:09:47.480> of 00:09:47.630 --> 00:09:47.640 align:start position:0% tokenization stemming and maybe a lot of 00:09:47.640 --> 00:09:49.310 align:start position:0% tokenization stemming and maybe a lot of other<00:09:47.880> Transformations<00:09:48.680> that<00:09:48.959> so<00:09:49.079> it's<00:09:49.160> on 00:09:49.310 --> 00:09:49.320 align:start position:0% other Transformations that so it's on 00:09:49.320 --> 00:09:51.990 align:start position:0% other Transformations that so it's on jvm<00:09:49.760> as<00:09:49.959> as<00:09:50.079> udfs<00:09:50.640> in<00:09:50.800> in<00:09:51.120> yeah<00:09:51.200> it's<00:09:51.320> all<00:09:51.480> jvm 00:09:51.990 --> 00:09:52.000 align:start position:0% jvm as as udfs in in yeah it's all jvm 00:09:52.000 --> 00:09:54.350 align:start position:0% jvm as as udfs in in yeah it's all jvm yes<00:09:52.160> it's<00:09:52.279> all<00:09:52.399> in<00:09:52.519> jvm<00:09:53.200> and<00:09:53.480> uh<00:09:53.600> some<00:09:53.720> of<00:09:53.920> those 00:09:54.350 --> 00:09:54.360 align:start position:0% yes it's all in jvm and uh some of those 00:09:54.360 --> 00:09:56.670 align:start position:0% yes it's all in jvm and uh some of those udfs<00:09:54.839> are<00:09:55.000> also<00:09:55.440> open-<00:09:55.760> sourced<00:09:56.079> in<00:09:56.440> uh<00:09:56.560> the 00:09:56.670 --> 00:09:56.680 align:start position:0% udfs are also open- sourced in uh the 00:09:56.680 --> 00:09:59.430 align:start position:0% udfs are also open- sourced in uh the project<00:09:56.959> called<00:09:57.120> Brick<00:09:57.440> House<00:09:57.880> okay<00:09:58.240> so<00:09:58.880> uh 00:09:59.430 --> 00:09:59.440 align:start position:0% project called Brick House okay so uh 00:09:59.440 --> 00:10:01.509 align:start position:0% project called Brick House okay so uh it's<00:09:59.600> actually<00:09:59.839> free<00:10:00.079> to<00:10:00.240> use<00:10:00.440> for<00:10:01.079> a<00:10:01.200> lot<00:10:01.360> of 00:10:01.509 --> 00:10:01.519 align:start position:0% it's actually free to use for a lot of 00:10:01.519 --> 00:10:05.910 align:start position:0% it's actually free to use for a lot of people<00:10:01.880> so<00:10:02.640> uh<00:10:03.600> using<00:10:04.600> so<00:10:05.079> they<00:10:05.160> have<00:10:05.320> UDF<00:10:05.760> such 00:10:05.910 --> 00:10:05.920 align:start position:0% people so uh using so they have UDF such 00:10:05.920 --> 00:10:08.269 align:start position:0% people so uh using so they have UDF such as<00:10:06.560> uh<00:10:06.680> where<00:10:06.880> you<00:10:06.959> can<00:10:07.160> pass<00:10:07.399> in<00:10:07.519> a<00:10:07.600> dictionary 00:10:08.269 --> 00:10:08.279 align:start position:0% as uh where you can pass in a dictionary 00:10:08.279 --> 00:10:10.110 align:start position:0% as uh where you can pass in a dictionary and<00:10:08.440> you<00:10:08.560> can<00:10:08.720> pass<00:10:08.880> in<00:10:09.000> a<00:10:09.079> piece<00:10:09.240> of<00:10:09.440> text<00:10:10.040> and 00:10:10.110 --> 00:10:10.120 align:start position:0% and you can pass in a piece of text and 00:10:10.120 --> 00:10:12.550 align:start position:0% and you can pass in a piece of text and then<00:10:10.279> the<00:10:10.399> text<00:10:10.720> would<00:10:11.160> efficiently<00:10:12.160> extract 00:10:12.550 --> 00:10:12.560 align:start position:0% then the text would efficiently extract 00:10:12.560 --> 00:10:13.949 align:start position:0% then the text would efficiently extract the<00:10:12.640> certain<00:10:12.920> entities<00:10:13.440> out<00:10:13.560> of<00:10:13.720> the 00:10:13.949 --> 00:10:13.959 align:start position:0% the certain entities out of the 00:10:13.959 --> 00:10:17.069 align:start position:0% the certain entities out of the dictionary<00:10:14.440> and<00:10:14.560> so<00:10:14.680> on<00:10:15.200> so<00:10:15.800> uh<00:10:16.519> uh<00:10:16.720> we<00:10:16.839> have<00:10:17.000> a 00:10:17.069 --> 00:10:17.079 align:start position:0% dictionary and so on so uh uh we have a 00:10:17.079 --> 00:10:19.509 align:start position:0% dictionary and so on so uh uh we have a bunch<00:10:17.279> of<00:10:17.399> these<00:10:17.560> udfs<00:10:18.160> which<00:10:18.279> we<00:10:18.480> then<00:10:19.000> use<00:10:19.320> to 00:10:19.509 --> 00:10:19.519 align:start position:0% bunch of these udfs which we then use to 00:10:19.519 --> 00:10:22.150 align:start position:0% bunch of these udfs which we then use to kind<00:10:19.680> of<00:10:20.480> uh<00:10:21.040> uh<00:10:21.240> extract<00:10:21.640> entities<00:10:22.000> from 00:10:22.150 --> 00:10:22.160 align:start position:0% kind of uh uh extract entities from 00:10:22.160 --> 00:10:23.829 align:start position:0% kind of uh uh extract entities from Freebase<00:10:22.880> we<00:10:23.000> use<00:10:23.200> certain<00:10:23.480> ranking 00:10:23.829 --> 00:10:23.839 align:start position:0% Freebase we use certain ranking 00:10:23.839 --> 00:10:25.670 align:start position:0% Freebase we use certain ranking techniques<00:10:24.240> and<00:10:24.399> relevance<00:10:24.839> techniques<00:10:25.480> to 00:10:25.670 --> 00:10:25.680 align:start position:0% techniques and relevance techniques to 00:10:25.680 --> 00:10:27.790 align:start position:0% techniques and relevance techniques to say<00:10:25.920> that<00:10:26.440> uh<00:10:26.760> these<00:10:26.920> are<00:10:27.040> the<00:10:27.160> most<00:10:27.399> important 00:10:27.790 --> 00:10:27.800 align:start position:0% say that uh these are the most important 00:10:27.800 --> 00:10:29.470 align:start position:0% say that uh these are the most important entities<00:10:28.519> because<00:10:28.880> there's<00:10:29.040> a<00:10:29.120> lot<00:10:29.240> of<00:10:29.399> of 00:10:29.470 --> 00:10:29.480 align:start position:0% entities because there's a lot of of 00:10:29.480 --> 00:10:31.110 align:start position:0% entities because there's a lot of of noise<00:10:29.640> in<00:10:29.760> free<00:10:29.959> Bay<00:10:30.240> as<00:10:30.320> well<00:10:30.519> so<00:10:30.640> you<00:10:30.760> would 00:10:31.110 --> 00:10:31.120 align:start position:0% noise in free Bay as well so you would 00:10:31.120 --> 00:10:32.750 align:start position:0% noise in free Bay as well so you would probably<00:10:31.399> pick<00:10:31.600> the<00:10:31.720> cream<00:10:32.120> do<00:10:32.279> some<00:10:32.440> sort<00:10:32.600> of 00:10:32.750 --> 00:10:32.760 align:start position:0% probably pick the cream do some sort of 00:10:32.760 --> 00:10:35.030 align:start position:0% probably pick the cream do some sort of ranking<00:10:33.399> pick<00:10:33.560> the<00:10:33.680> cream<00:10:34.000> maybe<00:10:34.399> top<00:10:34.800> half<00:10:34.920> a 00:10:35.030 --> 00:10:35.040 align:start position:0% ranking pick the cream maybe top half a 00:10:35.040 --> 00:10:37.350 align:start position:0% ranking pick the cream maybe top half a million<00:10:35.519> keywords<00:10:36.320> and<00:10:36.440> then<00:10:36.800> map<00:10:37.040> those<00:10:37.200> half 00:10:37.350 --> 00:10:37.360 align:start position:0% million keywords and then map those half 00:10:37.360 --> 00:10:39.509 align:start position:0% million keywords and then map those half a<00:10:37.480> million<00:10:37.720> keywords<00:10:38.120> into<00:10:38.360> 10,000<00:10:38.880> topics 00:10:39.509 --> 00:10:39.519 align:start position:0% a million keywords into 10,000 topics 00:10:39.519 --> 00:10:40.910 align:start position:0% a million keywords into 10,000 topics because<00:10:39.800> multiple<00:10:40.120> keywords<00:10:40.440> would<00:10:40.560> map<00:10:40.800> to 00:10:40.910 --> 00:10:40.920 align:start position:0% because multiple keywords would map to 00:10:40.920 --> 00:10:43.590 align:start position:0% because multiple keywords would map to the<00:10:41.040> same<00:10:41.279> topic<00:10:41.600> and<00:10:41.720> so<00:10:41.880> on<00:10:42.399> so<00:10:43.040> uh<00:10:43.320> mainly 00:10:43.590 --> 00:10:43.600 align:start position:0% the same topic and so on so uh mainly 00:10:43.600 --> 00:10:45.150 align:start position:0% the same topic and so on so uh mainly the<00:10:43.720> stack<00:10:44.000> that<00:10:44.120> we<00:10:44.240> use<00:10:44.399> is<00:10:44.560> still<00:10:44.800> hi<00:10:45.040> in 00:10:45.150 --> 00:10:45.160 align:start position:0% the stack that we use is still hi in 00:10:45.160 --> 00:10:48.069 align:start position:0% the stack that we use is still hi in udfs<00:10:45.800> for<00:10:46.079> building<00:10:46.440> these<00:10:46.600> models<00:10:47.560> uh<00:10:47.800> with<00:10:47.959> a 00:10:48.069 --> 00:10:48.079 align:start position:0% udfs for building these models uh with a 00:10:48.079 --> 00:10:50.550 align:start position:0% udfs for building these models uh with a little<00:10:48.279> bit<00:10:48.480> of<00:10:49.040> uh<00:10:49.560> python<00:10:49.920> to<00:10:50.079> just<00:10:50.279> kind<00:10:50.399> of 00:10:50.550 --> 00:10:50.560 align:start position:0% little bit of uh python to just kind of 00:10:50.560 --> 00:10:52.870 align:start position:0% little bit of uh python to just kind of do<00:10:50.839> the<00:10:51.600> uh<00:10:51.880> little<00:10:52.040> bit<00:10:52.279> whatever<00:10:52.639> a<00:10:52.720> little 00:10:52.870 --> 00:10:52.880 align:start position:0% do the uh little bit whatever a little 00:10:52.880 --> 00:10:54.230 align:start position:0% do the uh little bit whatever a little bit<00:10:53.000> of<00:10:53.279> machine<00:10:53.519> learning<00:10:53.760> training<00:10:54.079> is 00:10:54.230 --> 00:10:54.240 align:start position:0% bit of machine learning training is 00:10:54.240 --> 00:10:55.990 align:start position:0% bit of machine learning training is required<00:10:54.839> interesting<00:10:55.399> so<00:10:55.720> I<00:10:55.760> mean<00:10:55.880> that 00:10:55.990 --> 00:10:56.000 align:start position:0% required interesting so I mean that 00:10:56.000 --> 00:10:58.030 align:start position:0% required interesting so I mean that sounds<00:10:56.360> really<00:10:56.680> kind<00:10:56.800> of<00:10:57.079> a<00:10:57.240> scalable 00:10:58.030 --> 00:10:58.040 align:start position:0% sounds really kind of a scalable 00:10:58.040 --> 00:10:59.949 align:start position:0% sounds really kind of a scalable production<00:10:58.440> infrastructure<00:10:58.959> and<00:10:59.480> I<00:10:59.600> must<00:10:59.760> say 00:10:59.949 --> 00:10:59.959 align:start position:0% production infrastructure and I must say 00:10:59.959 --> 00:11:02.230 align:start position:0% production infrastructure and I must say that<00:11:00.079> you<00:11:00.200> know<00:11:00.680> B's<00:11:01.279> I<00:11:01.360> think<00:11:01.560> lead<00:11:01.800> is<00:11:01.959> Jerome 00:11:02.230 --> 00:11:02.240 align:start position:0% that you know B's I think lead is Jerome 00:11:02.240 --> 00:11:05.389 align:start position:0% that you know B's I think lead is Jerome Banks<00:11:02.639> who<00:11:02.760> is<00:11:02.920> our<00:11:03.480> common<00:11:04.480> former<00:11:04.880> colleague 00:11:05.389 --> 00:11:05.399 align:start position:0% Banks who is our common former colleague 00:11:05.399 --> 00:11:07.310 align:start position:0% Banks who is our common former colleague now<00:11:05.600> at<00:11:05.760> F<00:11:06.000> we<00:11:06.200> and<00:11:06.399> actually<00:11:06.720> we<00:11:06.800> are<00:11:06.920> planning 00:11:07.310 --> 00:11:07.320 align:start position:0% now at F we and actually we are planning 00:11:07.320 --> 00:11:10.870 align:start position:0% now at F we and actually we are planning to<00:11:07.639> present<00:11:08.519> some<00:11:08.720> of<00:11:08.959> the<00:11:09.240> open<00:11:09.720> sourcing<00:11:10.720> uh 00:11:10.870 --> 00:11:10.880 align:start position:0% to present some of the open sourcing uh 00:11:10.880 --> 00:11:13.550 align:start position:0% to present some of the open sourcing uh stages<00:11:11.279> of<00:11:11.440> that<00:11:11.680> project<00:11:12.000> and<00:11:12.200> some<00:11:12.480> of<00:11:12.639> scups 00:11:13.550 --> 00:11:13.560 align:start position:0% stages of that project and some of scups 00:11:13.560 --> 00:11:16.069 align:start position:0% stages of that project and some of scups uh<00:11:13.720> coming<00:11:14.120> coming<00:11:14.320> forward<00:11:14.920> yeah<00:11:15.480> uh 00:11:16.069 --> 00:11:16.079 align:start position:0% uh coming coming forward yeah uh 00:11:16.079 --> 00:11:18.990 align:start position:0% uh coming coming forward yeah uh interesting<00:11:16.600> so<00:11:17.079> so<00:11:17.360> given<00:11:17.839> that<00:11:18.160> you<00:11:18.560> you<00:11:18.839> you 00:11:18.990 --> 00:11:19.000 align:start position:0% interesting so so given that you you you 00:11:19.000 --> 00:11:20.949 align:start position:0% interesting so so given that you you you kind<00:11:19.160> of<00:11:19.440> set<00:11:19.720> up<00:11:20.040> this<00:11:20.200> infrastructure<00:11:20.800> you 00:11:20.949 --> 00:11:20.959 align:start position:0% kind of set up this infrastructure you 00:11:20.959 --> 00:11:23.310 align:start position:0% kind of set up this infrastructure you toate<00:11:21.279> it<00:11:21.440> and<00:11:21.560> you<00:11:21.680> have<00:11:21.839> it<00:11:22.000> in<00:11:22.200> kind<00:11:22.320> of<00:11:22.519> nice 00:11:23.310 --> 00:11:23.320 align:start position:0% toate it and you have it in kind of nice 00:11:23.320 --> 00:11:25.110 align:start position:0% toate it and you have it in kind of nice you<00:11:23.399> know<00:11:23.560> ready<00:11:23.920> working<00:11:24.440> production<00:11:24.800> order 00:11:25.110 --> 00:11:25.120 align:start position:0% you know ready working production order 00:11:25.120 --> 00:11:26.829 align:start position:0% you know ready working production order and<00:11:25.200> you<00:11:25.279> can<00:11:25.399> iterate<00:11:25.720> an<00:11:25.880> experiment<00:11:26.760> what 00:11:26.829 --> 00:11:26.839 align:start position:0% and you can iterate an experiment what 00:11:26.839 --> 00:11:28.389 align:start position:0% and you can iterate an experiment what are<00:11:27.040> the<00:11:27.200> next<00:11:27.440> steps<00:11:27.800> where<00:11:27.920> do<00:11:28.040> you<00:11:28.120> see 00:11:28.389 --> 00:11:28.399 align:start position:0% are the next steps where do you see 00:11:28.399 --> 00:11:30.710 align:start position:0% are the next steps where do you see challenges<00:11:28.880> for<00:11:29.320> for<00:11:29.760> for<00:11:30.000> this<00:11:30.240> project<00:11:30.600> in 00:11:30.710 --> 00:11:30.720 align:start position:0% challenges for for for this project in 00:11:30.720 --> 00:11:32.910 align:start position:0% challenges for for for this project in the<00:11:30.839> future<00:11:31.720> yeah<00:11:31.920> I<00:11:32.000> think<00:11:32.399> uh<00:11:32.560> one<00:11:32.680> of<00:11:32.800> the 00:11:32.910 --> 00:11:32.920 align:start position:0% the future yeah I think uh one of the 00:11:32.920 --> 00:11:35.269 align:start position:0% the future yeah I think uh one of the most<00:11:33.519> exciting<00:11:34.000> things<00:11:34.320> for<00:11:34.519> me<00:11:34.760> is<00:11:35.120> uh 00:11:35.269 --> 00:11:35.279 align:start position:0% most exciting things for me is uh 00:11:35.279 --> 00:11:36.710 align:start position:0% most exciting things for me is uh actually<00:11:35.519> in<00:11:35.600> the<00:11:35.680> domain<00:11:35.920> of<00:11:36.040> deep<00:11:36.240> learning 00:11:36.710 --> 00:11:36.720 align:start position:0% actually in the domain of deep learning 00:11:36.720 --> 00:11:39.350 align:start position:0% actually in the domain of deep learning MH<00:11:37.160> uh<00:11:37.440> so<00:11:37.600> deep<00:11:37.800> learning<00:11:38.160> is<00:11:38.320> just<00:11:38.639> I<00:11:38.680> mean 00:11:39.350 --> 00:11:39.360 align:start position:0% MH uh so deep learning is just I mean 00:11:39.360 --> 00:11:41.110 align:start position:0% MH uh so deep learning is just I mean everybody's<00:11:39.800> talking<00:11:40.120> about<00:11:40.360> it<00:11:40.600> it's<00:11:40.959> it's 00:11:41.110 --> 00:11:41.120 align:start position:0% everybody's talking about it it's it's 00:11:41.120 --> 00:11:42.870 align:start position:0% everybody's talking about it it's it's kind<00:11:41.240> of<00:11:41.320> the<00:11:41.480> hot<00:11:41.920> area<00:11:42.160> of<00:11:42.240> machine<00:11:42.519> learning 00:11:42.870 --> 00:11:42.880 align:start position:0% kind of the hot area of machine learning 00:11:42.880 --> 00:11:45.430 align:start position:0% kind of the hot area of machine learning right<00:11:43.040> now<00:11:43.720> uh<00:11:43.800> the<00:11:43.920> challenge<00:11:44.399> there<00:11:44.600> is<00:11:44.800> that 00:11:45.430 --> 00:11:45.440 align:start position:0% right now uh the challenge there is that 00:11:45.440 --> 00:11:46.990 align:start position:0% right now uh the challenge there is that if<00:11:45.519> you<00:11:45.600> wanted<00:11:45.880> to<00:11:46.000> run<00:11:46.240> any<00:11:46.399> algorithm<00:11:46.880> on 00:11:46.990 --> 00:11:47.000 align:start position:0% if you wanted to run any algorithm on 00:11:47.000 --> 00:11:49.590 align:start position:0% if you wanted to run any algorithm on deep<00:11:47.200> learning<00:11:47.639> you<00:11:48.360> it's<00:11:49.160> I<00:11:49.200> mean<00:11:49.399> it's 00:11:49.590 --> 00:11:49.600 align:start position:0% deep learning you it's I mean it's 00:11:49.600 --> 00:11:50.990 align:start position:0% deep learning you it's I mean it's debatable<00:11:50.120> whether<00:11:50.320> you<00:11:50.440> can<00:11:50.560> do<00:11:50.680> it<00:11:50.800> on<00:11:50.920> the 00:11:50.990 --> 00:11:51.000 align:start position:0% debatable whether you can do it on the 00:11:51.000 --> 00:11:53.030 align:start position:0% debatable whether you can do it on the cloud<00:11:51.279> and<00:11:51.480> on<00:11:51.600> the<00:11:51.680> jvm<00:11:52.399> uh<00:11:52.639> right<00:11:52.760> now<00:11:52.920> the 00:11:53.030 --> 00:11:53.040 align:start position:0% cloud and on the jvm uh right now the 00:11:53.040 --> 00:11:55.629 align:start position:0% cloud and on the jvm uh right now the best<00:11:53.760> uh<00:11:54.120> approaches<00:11:54.519> to<00:11:54.680> that<00:11:54.880> have<00:11:55.040> been<00:11:55.480> to 00:11:55.629 --> 00:11:55.639 align:start position:0% best uh approaches to that have been to 00:11:55.639 --> 00:11:58.990 align:start position:0% best uh approaches to that have been to run<00:11:55.839> it<00:11:56.000> on<00:11:56.680> uh<00:11:57.160> gpus<00:11:58.160> uh<00:11:58.440> now<00:11:58.600> if<00:11:58.680> you<00:11:58.800> wanted 00:11:58.990 --> 00:11:59.000 align:start position:0% run it on uh gpus uh now if you wanted 00:11:59.000 --> 00:12:01.110 align:start position:0% run it on uh gpus uh now if you wanted to<00:11:59.240> scale<00:11:59.519> gpus<00:12:00.320> but<00:12:00.560> now<00:12:00.680> you're<00:12:00.839> using 00:12:01.110 --> 00:12:01.120 align:start position:0% to scale gpus but now you're using 00:12:01.120 --> 00:12:02.949 align:start position:0% to scale gpus but now you're using something<00:12:01.399> like<00:12:01.519> the<00:12:01.639> Hado<00:12:02.000> stack<00:12:02.680> then<00:12:02.839> how 00:12:02.949 --> 00:12:02.959 align:start position:0% something like the Hado stack then how 00:12:02.959 --> 00:12:04.430 align:start position:0% something like the Hado stack then how you<00:12:03.079> going<00:12:03.200> to<00:12:03.360> really<00:12:03.560> go<00:12:03.720> about<00:12:03.920> it<00:12:04.079> so<00:12:04.240> I 00:12:04.430 --> 00:12:04.440 align:start position:0% you going to really go about it so I 00:12:04.440 --> 00:12:06.509 align:start position:0% you going to really go about it so I think<00:12:04.880> uh<00:12:05.440> there<00:12:05.519> are<00:12:05.680> certain<00:12:05.959> challenges 00:12:06.509 --> 00:12:06.519 align:start position:0% think uh there are certain challenges 00:12:06.519 --> 00:12:09.350 align:start position:0% think uh there are certain challenges there<00:12:06.839> that<00:12:07.240> uh<00:12:07.800> we'll<00:12:08.040> probably<00:12:08.480> be<00:12:08.880> I<00:12:08.959> mean 00:12:09.350 --> 00:12:09.360 align:start position:0% there that uh we'll probably be I mean 00:12:09.360 --> 00:12:11.829 align:start position:0% there that uh we'll probably be I mean I'll<00:12:09.519> be<00:12:09.680> keeping<00:12:09.920> my<00:12:10.360> eye<00:12:10.519> out<00:12:10.720> on<00:12:11.519> uh<00:12:11.600> in<00:12:11.720> the 00:12:11.829 --> 00:12:11.839 align:start position:0% I'll be keeping my eye out on uh in the 00:12:11.839 --> 00:12:13.150 align:start position:0% I'll be keeping my eye out on uh in the next<00:12:12.040> couple<00:12:12.240> of<00:12:12.360> years<00:12:12.519> to<00:12:12.639> see<00:12:12.839> how<00:12:13.000> that 00:12:13.150 --> 00:12:13.160 align:start position:0% next couple of years to see how that 00:12:13.160 --> 00:12:15.670 align:start position:0% next couple of years to see how that plans<00:12:13.480> out<00:12:13.800> but<00:12:14.240> uh<00:12:14.800> uh<00:12:14.959> that<00:12:15.120> is<00:12:15.279> a<00:12:15.440> very 00:12:15.670 --> 00:12:15.680 align:start position:0% plans out but uh uh that is a very 00:12:15.680 --> 00:12:18.590 align:start position:0% plans out but uh uh that is a very exciting<00:12:16.320> area<00:12:16.639> of<00:12:17.079> research<00:12:17.519> because<00:12:18.360> uh<00:12:18.480> if 00:12:18.590 --> 00:12:18.600 align:start position:0% exciting area of research because uh if 00:12:18.600 --> 00:12:20.389 align:start position:0% exciting area of research because uh if you<00:12:18.680> are<00:12:18.800> able<00:12:19.000> to<00:12:19.079> do<00:12:19.279> unsupervised<00:12:20.079> learning 00:12:20.389 --> 00:12:20.399 align:start position:0% you are able to do unsupervised learning 00:12:20.399 --> 00:12:22.269 align:start position:0% you are able to do unsupervised learning using<00:12:20.920> uh<00:12:21.040> deep<00:12:21.279> learning<00:12:21.600> on<00:12:21.800> like<00:12:21.959> large 00:12:22.269 --> 00:12:22.279 align:start position:0% using uh deep learning on like large 00:12:22.279 --> 00:12:24.949 align:start position:0% using uh deep learning on like large corpuses<00:12:23.040> then<00:12:23.839> uh<00:12:23.959> you<00:12:24.120> pretty<00:12:24.360> much<00:12:24.639> have 00:12:24.949 --> 00:12:24.959 align:start position:0% corpuses then uh you pretty much have 00:12:24.959 --> 00:12:28.470 align:start position:0% corpuses then uh you pretty much have like<00:12:25.480> uh<00:12:26.120> a<00:12:26.320> huge<00:12:26.880> step<00:12:27.199> ahead<00:12:27.560> than<00:12:27.839> what 00:12:28.470 --> 00:12:28.480 align:start position:0% like uh a huge step ahead than what 00:12:28.480 --> 00:12:31.470 align:start position:0% like uh a huge step ahead than what everybody<00:12:28.920> has<00:12:29.040> is<00:12:29.240> that<00:12:29.399> currently<00:12:29.839> on<00:12:30.839> uh<00:12:31.320> in 00:12:31.470 --> 00:12:31.480 align:start position:0% everybody has is that currently on uh in 00:12:31.480 --> 00:12:33.550 align:start position:0% everybody has is that currently on uh in in<00:12:31.560> terms<00:12:31.800> of<00:12:31.920> production<00:12:32.320> systems<00:12:33.120> you<00:12:33.240> know 00:12:33.550 --> 00:12:33.560 align:start position:0% in terms of production systems you know 00:12:33.560 --> 00:12:34.629 align:start position:0% in terms of production systems you know it's<00:12:33.680> interesting<00:12:34.079> you<00:12:34.160> mention<00:12:34.440> this 00:12:34.629 --> 00:12:34.639 align:start position:0% it's interesting you mention this 00:12:34.639 --> 00:12:36.670 align:start position:0% it's interesting you mention this because<00:12:34.839> you<00:12:34.920> know<00:12:35.079> we<00:12:35.199> have<00:12:35.360> this<00:12:35.800> uh<00:12:36.519> uh 00:12:36.670 --> 00:12:36.680 align:start position:0% because you know we have this uh uh 00:12:36.680 --> 00:12:38.629 align:start position:0% because you know we have this uh uh conference<00:12:37.199> coming<00:12:37.440> up<00:12:37.639> text<00:12:37.920> by<00:12:38.079> the<00:12:38.160> bay<00:12:38.440> and 00:12:38.629 --> 00:12:38.639 align:start position:0% conference coming up text by the bay and 00:12:38.639 --> 00:12:40.310 align:start position:0% conference coming up text by the bay and one<00:12:38.760> of<00:12:38.880> the<00:12:39.000> talks<00:12:39.279> they<00:12:39.519> actually<00:12:39.800> Dearing 00:12:40.310 --> 00:12:40.320 align:start position:0% one of the talks they actually Dearing 00:12:40.320 --> 00:12:42.790 align:start position:0% one of the talks they actually Dearing with<00:12:40.480> deing<00:12:40.880> for<00:12:41.040> J<00:12:41.639> on<00:12:41.800> spark<00:12:42.160> and<00:12:42.320> that<00:12:42.480> tool 00:12:42.790 --> 00:12:42.800 align:start position:0% with deing for J on spark and that tool 00:12:42.800 --> 00:12:45.470 align:start position:0% with deing for J on spark and that tool kit<00:12:43.079> actually<00:12:43.800> uses<00:12:44.199> gpus<00:12:44.800> right<00:12:44.920> so<00:12:45.079> because 00:12:45.470 --> 00:12:45.480 align:start position:0% kit actually uses gpus right so because 00:12:45.480 --> 00:12:47.990 align:start position:0% kit actually uses gpus right so because you<00:12:45.800> you<00:12:45.920> can<00:12:46.399> there<00:12:46.519> is<00:12:46.639> a<00:12:47.040> linear<00:12:47.480> algebra 00:12:47.990 --> 00:12:48.000 align:start position:0% you you can there is a linear algebra 00:12:48.000 --> 00:12:51.790 align:start position:0% you you can there is a linear algebra implementation<00:12:48.680> nd4j<00:12:49.399> which<00:12:49.600> can<00:12:50.160> use<00:12:50.920> Cuda 00:12:51.790 --> 00:12:51.800 align:start position:0% implementation nd4j which can use Cuda 00:12:51.800 --> 00:12:55.269 align:start position:0% implementation nd4j which can use Cuda and<00:12:51.959> and<00:12:52.120> gpus<00:12:52.639> so<00:12:53.160> hopefully<00:12:54.160> uh<00:12:54.959> we<00:12:55.160> make 00:12:55.269 --> 00:12:55.279 align:start position:0% and and gpus so hopefully uh we make 00:12:55.279 --> 00:12:56.990 align:start position:0% and and gpus so hopefully uh we make some<00:12:55.440> advances<00:12:56.000> there<00:12:56.240> right<00:12:56.399> and<00:12:56.880> I've 00:12:56.990 --> 00:12:57.000 align:start position:0% some advances there right and I've 00:12:57.000 --> 00:12:58.509 align:start position:0% some advances there right and I've looked<00:12:57.160> into<00:12:57.360> de<00:12:57.720> learning<00:12:57.959> for<00:12:58.240> it<00:12:58.320> looks 00:12:58.509 --> 00:12:58.519 align:start position:0% looked into de learning for it looks 00:12:58.519 --> 00:13:00.150 align:start position:0% looked into de learning for it looks very<00:12:58.680> interesting 00:13:00.150 --> 00:13:00.160 align:start position:0% very interesting 00:13:00.160 --> 00:13:02.069 align:start position:0% very interesting Park<00:13:00.399> too<00:13:00.680> right<00:13:00.839> so<00:13:01.199> hopefully<00:13:01.760> all<00:13:01.920> this 00:13:02.069 --> 00:13:02.079 align:start position:0% Park too right so hopefully all this 00:13:02.079 --> 00:13:03.550 align:start position:0% Park too right so hopefully all this stuff<00:13:02.240> will<00:13:02.399> come<00:13:02.680> together<00:13:02.920> and<00:13:03.079> I<00:13:03.160> think 00:13:03.550 --> 00:13:03.560 align:start position:0% stuff will come together and I think 00:13:03.560 --> 00:13:05.790 align:start position:0% stuff will come together and I think actually<00:13:03.959> it<00:13:04.240> a<00:13:04.399> great<00:13:04.639> application<00:13:05.639> and 00:13:05.790 --> 00:13:05.800 align:start position:0% actually it a great application and 00:13:05.800 --> 00:13:08.350 align:start position:0% actually it a great application and hopefully<00:13:06.240> you<00:13:06.360> know<00:13:07.040> uh<00:13:07.279> we'll<00:13:07.480> see<00:13:08.120> what's 00:13:08.350 --> 00:13:08.360 align:start position:0% hopefully you know uh we'll see what's 00:13:08.360 --> 00:13:09.910 align:start position:0% hopefully you know uh we'll see what's next<00:13:08.600> and<00:13:08.760> I<00:13:08.839> mean<00:13:09.240> I<00:13:09.360> think<00:13:09.480> the<00:13:09.600> committee 00:13:09.910 --> 00:13:09.920 align:start position:0% next and I mean I think the committee 00:13:09.920 --> 00:13:12.150 align:start position:0% next and I mean I think the committee will<00:13:10.040> be<00:13:10.160> happy<00:13:10.440> to<00:13:10.959> to<00:13:11.120> work<00:13:11.279> on<00:13:11.399> this<00:13:11.600> stuff 00:13:12.150 --> 00:13:12.160 align:start position:0% will be happy to to work on this stuff 00:13:12.160 --> 00:13:14.189 align:start position:0% will be happy to to work on this stuff yeah<00:13:12.480> well<00:13:12.680> thanks<00:13:13.120> for<00:13:13.279> sharing<00:13:13.639> us<00:13:14.079> we're 00:13:14.189 --> 00:13:14.199 align:start position:0% yeah well thanks for sharing us we're 00:13:14.199 --> 00:13:15.470 align:start position:0% yeah well thanks for sharing us we're looking<00:13:14.360> forward<00:13:14.600> to<00:13:14.720> your<00:13:14.920> talk<00:13:15.199> thank<00:13:15.320> you 00:13:15.470 --> 00:13:15.480 align:start position:0% looking forward to your talk thank you 00:13:15.480 --> 00:13:28.110 align:start position:0% looking forward to your talk thank you thanks<00:13:15.720> so<00:13:15.839> much<00:13:16.320> yeah<00:13:16.480> all<00:13:16.600> right 00:13:28.110 --> 00:13:28.120 align:start position:0% 00:13:28.120 --> 00:13:31.120 align:start position:0% thanks