22nd International Joint Conference on Artificial Intelligence (IJCAI), Barcelona 2011

Open Information Extraction at Web Scale

author: Oren Etzioni, Turing Center, University of Washington
published: Aug. 23, 2011, recorded: July 2011, views: 1031

Slides

Slides
0:00	Open Information Extraction at Web Scale
0:13	KnowItAll Group (2003 - ?)
0:31	Les Valiant (Turing Award 2011)
1:04	Knowledge Acquisition Bottleneck
4:16	What is Machine Reading?
4:57	More Pragmatic Motivation: Information Overload
5:34	Paradigm Shift: from retrieval to reading
6:54	RevMiner (Huang, Etzioni, Zettlemoyer)
8:03	revminer - 1
8:09	revminer - 2
10:04	revminer - 1
10:07	revminer - 2
10:14	Outline - 1
11:46	Information Extraction (IE)
12:26	Context → clues
13:05	How to Scale IE?
14:41	Learned Extraction Clues
15:41	No.
15:55	Critique of IE=supervised learning
16:58	Semi-Supervised Learning
18:35	Outline - 2
19:07	Open IE (Banko, IJCAI ’07; ACL ‘08)
19:43	Open versus Traditional IE
20:43	TextRunner
21:24	Relation Extraction in TextRunner
22:50	TextRunner Architecture
23:40	Two Types of Extraction Errors
24:45	How to Filter Unsound Extractions?
26:02	Combinatorial Model (Downey, IJCAI ’05, AIJ ‘10)
26:49	Key Ideas in TextRunner
27:36	Error Analysis of TextRunner Relations
28:40	ReVerb (Fader, EMNLP ’11; Etzioni et al., IJCAI ‘11)
30:54	ReVerb Refinement
31:39	Sample of ReVerb Relations
31:52	Number of Relations
32:18	ReVerb versus TextRunner
34:09	Demo
34:52	ReVerb Search - 1
35:26	ReVerb Search - 2
38:44	Have we made progress towards Machine Reading?
39:00	Extractions as basis for Inference
40:52	Synonyms (Mars = Red Planet)
42:42	Argument Typing
42:58	Text → Argument Types
43:43	TextRunner Extractions
44:09	LinkLDA
44:33	Demo of LDA-SP
45:55	Open IE Lessons
46:27	Conclusions/Speculations

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.