NATO Advanced Study Institute on Mining Massive Data Sets for Security

Learning to Extract Security-related Event Information from Large News Collections

author: Hristo Tanev, Joint Research Centre

Description

Automatic Event Extraction from texts emerges as an im- portant and complex text mining task. Its goal is to detect description of events of a speci¯c type described in the text. For each event the Event Extraction system is expected to ¯nd the time, the location, the participants in this event and their roles, as well as other related circum- stances. In this talk we present a Machine Learning approach for learning of information extraction patterns, a method for semi-automatic lexical acquisition, and an information aggregation strategy implemented in a working prototype nexus which detects automatically security related events in clusters of news articles.

You might be experiencing some problems with Your Video player.

*Slides*
0:00	Learning to Extract Security-related Event Information from Large News Collections
0:29	Outline
1:21	Outline - Event Extraction overview and background
1:24	Automatic Event Extraction
2:47	Event description example (1)
3:01	Event description example (2)
3:39	Event structure and relations between events
4:36	Problems in front of Automatic Event Extraction
5:55	State of the art (1)
6:57	State of the art (2)
7:34	Machine Learning Approaches for Event Extraction
9:21	Outline - Overview of the Event Extraction system NEXUS, profiled for the security domain
9:38	Security-related event extraction, the NEXUS system
9:46	Motivation
10:12	European Media Monitor (EMM)
11:17	NEXUS – a prototype system for automatic event extraction from news
11:37	NEXUS processing chain
13:00	Main topics related to NEXUS
13:26	Outline - Pattern learning
13:33	Learning patterns
14:02	Learning patterns- overview
14:34	Learning patterns- an example
17:08	Learning patterns (1)
18:35	Learning patterns (2)
20:14	Learning patterns (3)
21:23	Learning patterns (2)
21:27	Learning patterns (3)
21:53	Learning patterns (4)
22:57	Learning patterns (5)
23:22	Learning patterns –context
23:35	Learning patterns (4)
23:57	Learning patterns –context
24:40	Learning patterns - context entropy
24:46	Learning patterns –context
25:03	Learning patterns - context entropy
25:29	Learning patterns – partial order of the pattern candidates
26:06	Learning patterns – context entropy
26:18	Learning patterns – partial order of the pattern candidates
26:36	Learning patterns – context entropy
27:12	Learning patterns – partial order of the pattern candidates
27:49	Learning patterns – Local Maximum of the Context Entropy
28:13	Expanding the set of patterns
29:38	Pattern library
30:05	Outline - Lexicon acquisition
30:35	- Questions
32:43	Outline - Lexicon acquisition
33:12	Semi-automatic construction of lexicons
33:58	Acquisition of a lexicon of weapons (1)
35:19	Acquisition of a lexicon of weapons (2)
36:01	Acquisition of a lexicon of weapons (3)
36:38	Acquisition of a lexicon of weapons (4)
38:48	Expanding the lexicon
40:15	Classification of the weapons (1)
40:31	Classification of the weapons (2)
40:47	Event classification lexicon
41:03	Event classification lexicon– event type labels
41:46	Event classification lexicon
42:12	Event classification
42:36	Outline - Event aggregation
42:49	NEXUS processing chain
43:43	Event aggregation (1)
43:57	Event aggregation (2)
44:21	NEXUS processing chain
45:09	Event aggregation (2)
45:15	Event aggregation. Tracking down the main event of the cluster
45:48	Event aggregation (2)
46:28	Event aggregation. Resolving role ambiguities
46:59	Event aggregation. Event geocoding
47:34	Example for Event Extraction
47:51	Example. Defining number of killed, wounded, and kidnapped (1)
48:47	Example. Defining number of killed, wounded, and kidnapped (2)
48:57	Example. Finding perpetrators
49:21	Example. Weapon detection
49:30	Example. Type detection (1)
50:07	Example. Type detection (2)
50:09	Example. Event geocoding
50:14	Example. Detect victim descriptions
50:22	Example. The final event frame
50:41	Outline - Evaluation and applications
50:43	Preliminary precision evaluation
51:35	Applications (1)
51:47	Applications (2)
51:55	Applications (3)
58:18	Outline - Conclusions and future work
58:21	Conclusions
58:59	Future directions

Lecture rating

People found this lecture:
Worth seeing
because it is:
Valuable and informative
Well presented
Easily understandable
Acceptably recorded
You need to login to cast your vote.

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.