NIPS ˙08 Workshop: Beyond Search - Computational Intelligence for the Web

Beyond the Semantic Web

author: Doug Lenat, Cycorp Inc.
published: Dec. 20, 2008, recorded: December 2008, views: 965

Slides

Slides
0:00	Beyond the Semantic Web
0:12	Beyond the Semantic Web
0:24	There and back again
1:03	My entry into AI (1)
1:25	My entry into AI (2)
1:28	My entry into AI (3)
2:43	My entry into AI (4)
3:21	My entry into AI (5)
3:35	My entry into AI (6)
3:43	My entry into AI (7)
3:50	My entry into AI (8)
3:58	My entry into AI (9)
4:25	My entry into AI (10)
4:40	My entry into AI (11)
4:43	Progress
5:37	publication
6:00	AM (1)
6:21	AM (2)
6:45	AM (3)
6:57	AM (4)
7:15	AM (5)
7:35	AM (6)
8:16	AM (7)
8:26	AM (8)
9:03	AM (9)
9:04	AM (10)
9:06	AM (11)
9:15	AM (10)
9:44	AM (11)
9:51	AM (12)
10:05	AM (13)
10:33	AM (14)
10:36	AM (15)
10:42	AM (16)
10:50	Artificial Inteligence
11:32	AM-conclusion
11:35	What’s in a task on the AM agenda
11:52	What’s in an AM heuristic?
12:03	How did AM “discover” numbers, +…? (1)
12:53	How did AM “discover” numbers, +…? (2)
12:58	AM - conclusion (2)
13:13	EURISKO (1)
13:32	EURISKO (2)
13:42	EURISKO (3)
13:43	EURISKO (4)
15:01	EURISKO (5)
20:15	EURISKO (6)
20:16	EURISKO (7)
20:20	EURISKO (8)
20:34	So if you’re going to mutate heuristics…
20:36	When will our programs ever learn(in “rich” domains unlike math/games)?
23:34	progress to Cyc
23:56	CYC (1)
24:07	CYC (2)
24:11	CYC Knowledge base
24:21	Cyc knowledge Base - focused
24:46	CYC Knowledge base
25:21	Cyc knowledge Base - focused
25:29	Relations Between an Event and its Participants
25:58	Propositional Attitudes
26:13	In In Our Geospatial Ontology (1)
26:34	In In Our Geospatial Ontology (2)
26:46	In In Our Geospatial Ontology (3)
26:47	In In Our Geospatial Ontology (4)
26:50	In In Our Geospatial Ontology (5)
26:53	In In Our Geospatial Ontology (6)
26:56	Concepts are densely related
27:07	Example
27:36	Syntactic Power
27:54	1972-89: Forced to more and more expressive representation languages: (1)
28:40	1972-89: Forced to more and more expressive representation languages: (2)
29:05	1972-89: Forced to more and more expressive representation languages: (3)
29:10	For the EL, use as expressive a language as is called for.
29:58	Power (1)
29:59	Power (2)
30:17	45th’s Space Wing Hurricane Preparedness
30:52	Performance:Subtheory: disjointWith
31:33	Inference is Fast & Trainable
31:40	opencyc.org
31:54	Cyc Mission
32:47	Building Cyc qua Engineering Task
35:08	Manual Knowledge Entry
37:51	Modes of Acquisition
38:14	Ambitious Approach: General Automated Interview
38:27	Ambition is good (1)
38:36	Ambition is good (2)
38:39	Example
38:44	USE of Acquired Knowledge
39:08	The more Reasonable your data is, The cleverer things Cyc can help with (1)
39:14	The more Reasonable your data is, The cleverer things Cyc can help with (2)
39:49	Intelligent Search (1)
39:51	Intelligent Search (2)
40:14	Intelligent Search (3)
40:24	Intelligent Search (4)
40:25	Intelligent Search (5)
40:28	Intelligent Search (6)
40:30	Intelligent Search (7)
40:35	Intelligent Search (8)
40:36	Intelligent Search - results
40:45	Intelligent Search-2
40:47	Intelligent Search-2 results
40:54	more related results
41:15	Cyc possibilities
41:18	Cyc: Semantic research Assistant (1)
41:22	Cyc: Semantic Research Assistant (2)
41:54	Cyc Search (1)
42:02	Cyc Search (2)
42:04	Cyc Search (3)
42:08	Cyc Search (4)
42:09	Cyc Search (5)
42:15	Cyc Search (6)
42:18	Cyc Search (7)
42:19	Cyc Search (8)
42:23	Cyc Search (9)
42:25	Cyc Search (10)
42:28	Cyc Search (11)
42:30	Cyc Search (12)
42:31	Some opportunities for ML
43:04	Document Tagging
43:09	Knowledge-based disambiguation (1)
43:23	Disambiguation Rules: "Jet"
43:33	Knowledge-based disambiguation (2)
43:43	Knowledge-based disambiguation (3)
43:57	Rule Induction
43:59	The Induction pipeline
44:00	Early results
44:02	Sample rules produced
44:07	Early results
44:16	Sample rules produced
44:25	Early ML Experimetns
44:28	Markov Logic Work
44:47	Cyc Knowledge base
44:52	AKA By Shallow Fishing (1)
45:11	AKA By Shallow Fishing (2)
45:13	AKA By Shallow Fishing (3)
45:14	AKA By Shallow Fishing (4)
45:16	AKA By Shallow Fishing (5)
45:20	AKA By Shallow Fishing (6)
45:31	Example
45:56	AKA By Shallow Fishing (7)
45:58	AKA By Shallow Fishing (8)
46:07	AKA By Shallow Fishing (9)
46:14	AKA By Shallow Fishing (10)
46:20	AKA By Shallow Fishing (11)
46:20	Analysis of Errors from that expt.
46:23	another search example
46:25	AKA By Shallow Fishing (12)
46:35	How to help Cyc (1)
47:00	How to help Cyc (2)
47:07	How to help Cyc (3)
47:09	How to help Cyc (4)
47:10	How to help Cyc (5)
47:12	How to help Cyc (6)
47:14	Verification by Volunteers
47:16	FACTory
47:21	Knowledge Acquisition Goals
47:25	Early results
47:27	Learning facts by search
47:34	Parsing results
47:37	verify. KB and exact search
47:38	Typical query for outcomes study
47:40	Typical Query
47:43	Typical query for outcomes study
47:53	The Cyc Analytic Environment (3)
47:58	The Cyc Analytic Environment (4)
47:59	The Cyc Analytic Environment (6)
48:14	Google search
48:22	The Analyst's Knowledge Base
48:24	the chain of events (1)
48:38	the chain of events (2)
48:40	the chain of events (3)
48:43	the chain of events (4)
48:45	the chain of events (5)
48:47	the chain of events (6)
48:49	the chain of events (7)
48:52	the chain of events (8)
48:55	the chain of events (9)
49:11	outcome
49:15	Versions of Cyc
49:33	Acquring and Using Cyc
49:38	Converting Semantic Meta-Knowledge into Inductive Bias
49:41	Example
49:43	A solution that scales linearly (1)
49:45	A solution that scales linearly (2)
49:47	another example of use
49:52	The UO mostly impacts efficiency So where is the source of power? (2)
49:57	Beyond the semantic web: What needs to be shared?
50:21	Beyond the semantic web (1)
50:51	Beyond the Semantic Web (2)
51:14	Beyond the Semantic Web (3)
52:11	opencyc.org
52:27	- questions

Report a problem or upload files

If you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.

Lecture popularity: You need to login to cast your vote.

Description

There are multiple sources of power available for forming and propelling automobiles; analogously, there are several sources of power for forming and propelling thoughts. Besides the neural ones you're most familiar with, and the Semantic Web ones that have received the lion's share of hype in recent years, there are some additional ones that we are tapping into with some success. These deep semantic representations and operations are able to produce useful and in cases even novel conclusions requiring induction, abduction, and analogy, as well as deductive reasoning. I will illustrate this with case examples from recent Cyc applications, including terrorism scenario generation for intelligence analysts and ad hoc clinical trial question answering for medical researchers.