Empirical Methods in an Open Domain Question Answering System
1 online resource (37 pages) : PDF
University of North Carolina at Charlotte
Answering questions effectively involves a complex interconnected set of linguistic and statistical analysis tools, which can be difficult to investigate and evaluate on their own. To evaluate any one of them, I developed two composable, recursive linguistic annotation pipelines and many segments associated therewith in order to facilitate the generation, transformation, analysis and ranking of answers to open domain English questions. The resulting pipeline is to the author’s knowledge the first open-source analog to the Watson system, and has reached 41% precision in the first rank when answering trivia questions from Jeopardy!.
COMPUTATIONAL LINGUISTICSMACHINE LEARNINGNATURAL LANGUAGE PROCESSINGQUESTION ANSWERING
Ras, ZbigniewCukic, Bojan
Thesis (M.S.)--University of North Carolina at Charlotte, 2015.
This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s). For additional information, see http://rightsstatements.org/page/InC/1.0/.
Copyright is held by the author unless otherwise indicated.