1 00:00:00,000 --> 00:00:06,095 So let's look at the vision that Tim Berners-Lee [UNKNOWN] way back in 2000 2 00:00:06,095 --> 00:00:13,134 called the semantic web, which essentially puts together some of the ideas that we 3 00:00:13,134 --> 00:00:03,241 have been talking in the past few minutes. So to answer a, a question or query like , 4 00:00:23,166 --> 00:00:30,067 who is the leader of USA, a semantic web system or web intelligent system 5 00:00:30,067 --> 00:00:36,968 incorporating semantics would proceed something like as follows: we might 6 00:00:36,968 --> 00:00:43,870 imagine a site a.com which collects facts by processing lots of web data, 7 00:00:43,870 --> 00:00:49,139 As we shall see later in this week how that's done. So we get facts like Obama is 8 00:00:49,139 --> 00:00:54,214 President of U.S.A., Vladimir Putin is president of Russia, Pranab Mukherjee is 9 00:00:54,214 --> 00:00:59,159 president of India, Manmohan Singh is Prime Minister of India, and many other 10 00:00:59,159 --> 00:01:02,542 facts about presidents, premiers, prime ministers, etc. 11 00:01:02,542 --> 00:01:07,357 Another site might be extracting information about who is leader of which 12 00:01:07,357 --> 00:01:12,496 country might figure out that Manmohan Singh is leader of India, Zuma is leader 13 00:01:12,496 --> 00:01:15,880 of South Africa, Putin is leader of Russia, and so on. 14 00:01:17,120 --> 00:01:24,489 A third site now might combine. Facts from a.com and b.com, and come to a 15 00:01:24,489 --> 00:01:31,634 conclusion using rule earning that with some degree of confidence that if x is the 16 00:01:31,634 --> 00:01:38,538 president of c, than x is leader of c. The process by which a bunch of facts is 17 00:01:38,538 --> 00:01:45,174 generalized to a rule using techniques like rule mining that we have seen earlier 18 00:01:45,174 --> 00:01:51,648 is called inductive reasoning, as opposed to deductive reasoning, which is normal 19 00:01:51,648 --> 00:01:56,100 logical inference. Inductive reasoning is almost always 20 00:01:56,680 --> 00:02:03,350 probabilistic to a certain extent. Using some of the techniques that we've 21 00:02:03,350 --> 00:02:10,198 already seen. Next normal deductive reasoning allows us 22 00:02:10,198 --> 00:02:15,980 to combine the rules and facts to arrive at. 23 00:02:16,840 --> 00:02:23,901 The fact that Obama is the leader of the U.S.A., which is then the answer to our 24 00:02:23,901 --> 00:02:28,050 query. Further, this new fact is then added back 25 00:02:28,050 --> 00:02:35,200 to the appropriate part of the semantic way of dealing with facts of this nature. 26 00:02:35,580 --> 00:02:42,340 Now this vision, is a powerful vision expressed more than a decade ago. 27 00:02:42,760 --> 00:02:51,067 It's not exactly been realized today, but, much of the technology needed, to express 28 00:02:51,067 --> 00:02:58,160 facts and rules, in a form that can be shared, across, different, systems. 29 00:02:58,700 --> 00:03:05,873 Using XML languages such as RDFS, RD, we just call it RDF schema, and OWL or the 30 00:03:05,873 --> 00:03:11,758 Web Ontology Language. That technology has been developed by the 31 00:03:11,758 --> 00:03:18,380 World Wide Web Foundation, where Tim Berners-Lee plays an important role. 32 00:03:20,760 --> 00:03:25,721 So the web of data and semantics is in principle possible. 33 00:03:25,721 --> 00:03:29,400 The question is, who is populating this web. 34 00:03:31,820 --> 00:03:35,935 Web scale inference is in some sense also possible, 35 00:03:35,935 --> 00:03:41,908 Not necessarily happening in exactly the same way as initially envisioned. 36 00:03:41,908 --> 00:03:47,157 But is happening. With efforts such as Google Squared, if 37 00:03:47,157 --> 00:03:53,270 you just figured this out from the web. It's essentially Google's attempt to 38 00:03:53,270 --> 00:03:58,740 extract lots of different facts from the wide, the world wide web. 39 00:03:59,060 --> 00:04:05,691 Wolfram Alpha is another recent search engine which relies on learning lots of 40 00:04:05,691 --> 00:04:11,232 facts about the world. And, of course, there's Watson which we've 41 00:04:11,232 --> 00:04:16,940 come across earlier , the IBM program that won the Jeopardy challenge. 42 00:04:16,940 --> 00:04:25,653 These are efforts which don't necessarily use techniques like OWL and semantic web 43 00:04:25,653 --> 00:04:30,151 technologies. Though they have a similar intent in 44 00:04:30,151 --> 00:04:36,247 spirit, which is essentially to learn facts from the web and be able to reason 45 00:04:36,247 --> 00:04:42,890 about those facts in a web intelligence system, as opposed to merely searching for 46 00:04:42,890 --> 00:04:46,798 web pages. So to summarize, the Symantec web vision 47 00:04:46,798 --> 00:04:53,126 is about a web of data and semantics. Which is shared so that one can have 48 00:04:53,126 --> 00:05:00,371 inference or reasoning at web scale. A bunch of technologies which is designed 49 00:05:00,371 --> 00:05:07,245 to enable this, RDF or Resource Description Framework as it's expansion is 50 00:05:07,245 --> 00:05:13,693 the Web Ontology Language and various variance of that, as we can see very soon. 51 00:05:13,693 --> 00:05:19,858 These are all technologies designed to enable this sharing of data and semantics 52 00:05:19,858 --> 00:05:23,640 across the web. At the same time, 53 00:05:23,640 --> 00:05:30,399 They are use to actually perform reasoning, has not necessarily proceeded 54 00:05:30,399 --> 00:05:34,591 in exactly the same way as originally envisioned. 55 00:05:34,591 --> 00:05:41,350 Google Squared, Wolfram Alpha, Watson do in fact reason using facts learned from 56 00:05:41,350 --> 00:05:46,740 the web, but not necessarily using the same technology backbone. 57 00:05:47,760 --> 00:05:55,644 We shall return to the semantic web and some efforts which are in fact learning 58 00:05:55,644 --> 00:06:04,187 facts in RDF and OWL form a little later. For the moment let's turn to resolution 59 00:06:04,187 --> 00:06:12,040 and logic and how such deductive reasoning might actually take place within a 60 00:06:12,040 --> 00:06:18,383 semantic web engine, regardless of the exact technology it uses.