Sign inSign up
Open Information Extraction at Web Scale thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Open Information Extraction at Web Scale

Published on Aug 23, 201116136 Views

Oren Etzioni

Research interests include: fundemental problems in the study of intelligence, Web search, Machine Reading, and Machine Learning.

IJCAI 2011 - Barcelona

Related categories

Artificial Intelligence

Chapter list

Open Information Extraction at Web Scale00:00
KnowItAll Group (2003 - ?)00:13
Les Valiant (Turing Award 2011)00:31
Knowledge Acquisition Bottleneck01:04
What is Machine Reading?04:16
More Pragmatic Motivation: Information Overload04:57
Paradigm Shift: from retrieval to reading05:34
RevMiner (Huang, Etzioni, Zettlemoyer)06:54
revminer - 108:03
revminer - 208:09
Outline - 110:14
Information Extraction (IE)11:46
Context → clues12:26
How to Scale IE?13:05
Learned Extraction Clues14:41
No.15:41
Critique of IE=supervised learning15:55
Semi-Supervised Learning16:58
Outline - 218:35
Open IE (Banko, IJCAI ’07; ACL ‘08)19:07
Open versus Traditional IE19:43
TextRunner20:43
Relation Extraction in TextRunner21:24
TextRunner Architecture22:50
Two Types of Extraction Errors23:40
How to Filter Unsound Extractions?24:45
Combinatorial Model (Downey, IJCAI ’05, AIJ ‘10)26:02
Key Ideas in TextRunner26:49
Error Analysis of TextRunner Relations27:36
ReVerb (Fader, EMNLP ’11; Etzioni et al., IJCAI ‘11)28:40
ReVerb Refinement30:54
Sample of ReVerb Relations31:39
Number of Relations31:52
ReVerb versus TextRunner32:18
Demo34:09
ReVerb Search - 134:52
ReVerb Search - 235:26
Have we made progress towards Machine Reading?38:44
Extractions as basis for Inference39:00
Synonyms (Mars = Red Planet)40:52
Argument Typing42:42
Text → Argument Types42:58
TextRunner Extractions43:43
LinkLDA44:09
Demo of LDA-SP44:33
Open IE Lessons45:55
Conclusions/Speculations46:27