Publication Date
Fall 2017
Degree Type
Master's Project
Degree Name
Master of Science (MS)
Department
Computer Science
Abstract
A question answer system takes queries from the user in natural language and returns a short concise answer which best fits the response to the question. This report discusses the integration and implementation of question answer systems for English and Hindi as part of the open source search engine Yioop. We have implemented a question answer system for English and Hindi, keeping in mind users who use these languages as their primary language. The user should be able to query a set of documents and should get the answers in the same language. English and Hindi are very different when it comes to language structure, characters etc. We have implemented the Question Answer System so that it supports localization and improved Part of Speech tagging performance by storing the lexicon in the database instead of a file based lexicon. We have implemented a brill tagger variant for Part of Speech tagging of Hindi phrases and grammar rules for triplet extraction. We also improve Yioop’s lexical data handling support by allowing the user to add named entities. Our improvements to Yioop were then evaluated by comparing the retrieved answers against a dataset of answers known to be true. The test data for the question answering system included creating 2 indexes, 1 each for English and Hindi. These were created by configuring Yioop to crawl 200,000 wikipedia pages for each crawl. The crawls were configured to be domain specific so that English index consists of pages restricted to English text and Hindi index is restricted to pages with Hindi text. We then used a set of 50 questions on the English and Hindi systems. We recored, Hindi system to have an accuracy of about 55% for simple factoid questions and English question answer system to have an accuracy of 63%.
Recommended Citation
Shenoy, Salil, "Improve and Implement an Open Source Question Answering System" (2017). Master's Projects. 577.
DOI: https://doi.org/10.31979/etd.8ybk-zkpa
https://scholarworks.sjsu.edu/etd_projects/577
Comments
A question answer system takes queries from the user in natural language and returns a short concise answer which best fits the response to the question. This report discusses the integration and implementation of question answer systems for English and Hindi as part of the open source search engine Yioop. We have implemented a question answer system for English and Hindi, keeping in mind users who use these languages as their primary language. The user should be able to query a set of documents and should get the answers in the same language. English and Hindi are very different when it comes to language structure, characters etc. We have implemented the Question Answer System so that it supports localization and improved Part of Speech tagging performance by storing the lexicon in the database instead of a file based lexicon. We have implemented a brill tagger variant for Part of Speech tagging of Hindi phrases and grammar rules for triplet extraction. We also improve Yioop’s lexical data handling support by allowing the user to add named entities. Our improvements to Yioop were then evaluated by comparing the retrieved answers against a dataset of answers known to be true. The test data for the question answering system included creating 2 indexes, 1 each for English and Hindi. These were created by configuring Yioop to crawl 200,000 wikipedia pages for each crawl. The crawls were configured to be domain specific so that English index consists of pages restricted to English text and Hindi index is restricted to pages with Hindi text. We then used a set of 50 questions on the English and Hindi systems. We recored, Hindi system to have an accuracy of about 55% for simple factoid questions and English question answer system to have an accuracy of 63%.