identifying entity relationships in news reports 27. january 2010 martin jačala, jozef tvarožek...
TRANSCRIPT
Identifying Entity Relationships in News Reports
27. January 2010
Martin Jačala, Jozef TvarožekFaculty of Informatics and Information TechnologySlovak University of Technology in Bratislava, Slovakia
Introduction
Analysis of text extracted from news reports Identification of persons, organizations, etc. Large amount of available data Providing constantly updated information The same person in various situations Revealing new, previously “hidden”
information Feedback of the community
27. January 2010
Method overview Text extracted from HTML
documents Part-of-speech tagging
HMM based
Entity identification Important phase Building corpora
Relationship analysis Rule based, input from previous layers
Presentation layer User friendly, accessible
27. January 2010
Results http://ktokoho.info
User interface
Relations between entities
Users can contribute
User modeling
Reusable data
Evaluation on corpus of articles written in Slovak language with 60% recall
27. January 2010