Show simple item record

dc.contributor.authorGhosh, Debanjan
dc.date.accessioned2016-04-05T16:15:50Z
dc.date.available2016-04-05T16:15:50Z
dc.date.issued2006
dc.identifier.isbn9780542628566
dc.identifier.isbn0542628562
dc.identifier.other304941232
dc.identifier.urihttp://hdl.handle.net/10477/49151
dc.description.abstractEvent detection and recognition is an important unsolved problem in information extraction. It is the problem of mining and classifying events from sentences. The ability to detect complex event patterns in data is limited by the intricacy of the data (here, simple text data) representation. In this work we are interested in extracting relations between features to detect events described in the sentences. Our data set pertains to various chapters of 9/11 commission report. We selected four types of events to detect; they are travel, contact, conflict and transaction. We did not focus on discovering relations among entities (typically named entities) here; rather we limited ourselves in finding relations between subject-verb-object (SVO) pairs or subject-verb-noun complements (SVC) that is the grammatical link-pair in sentences. We describe an event extraction or event labeling technique based on kernel methods. Kernel methods have become increasingly popular because of their ability to map arbitrary objects to a Euclidian feature space. (Abstract shortened by UMI.)
dc.languageEnglish
dc.sourceDissertations & Theses @ SUNY Buffalo,ProQuest Dissertations & Theses Global
dc.subjectApplied sciences
dc.titleSentence level event mining with kernels using support vector machines
dc.typeDissertation/Thesis


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record