SMS BASED FAQ RETRIEVAL USING HYBRID SIMILARITY MEASURE

Abstract

Mobile technology gave contribution to the progress of media of communication for example: chats, emails and short message services (SMS). The Popularity, utility and simplicity of SMSes is encouraging people to access information via SMSes, accessing information via internet creates hassle, it’s not necessary that internet connection is always available. So user can clarify their query, make complaint and get updates of result etc by sending SMS. Accessing information in such a manner makes information access very economic and easy for everybody from rural to metro city people. “FAQ retrieval” means there is corpora of frequently asked questions, and user sends a query in SMS language to retrieve some information. Such systems finds best match from FAQ corpora for given user defined query written in SMS language. The main problem in SMS language is the noise associated with it. Spelling mistakes, transliteration, phonetic spellings, abbreviations and short forms create difficulties in string matching. In proposed work, a novel approach has been presented by developing Hybrid similarities which evaluates similarity scores with the questions in the corpus for SMS query. In this way, we can further improve the accuracy of the SMS based FAQ system significantly by refining the results of the system using different hybrid similarity scores.