Language Modeling Approach to Retrieval for SMS and FAQ Matching

Short Messaging service popularly known as “SMS” has seen growth due to the growth in Mobile phone users. A mobile phone is considered as a cheap and easy device for communication. It is also used as a source to acquire and spread information. SMS based FAQ Retrieval task proposed in FIRE 2011 aims to provide the required information from frequently asked questions (FAQs). Challenge is to find a question from corpora of FAQs that best answers/matches with the SMS query. But, SMS queries are noisy as users tend to compress text by omitting letters, using slang, etc. This is observed due to a cap on the length of messages (160 characters constitute one SMS), lack of screen space (which makes reading large amounts of text difficult). In this paper, we propose a method using language modeling approach to match noisy SMS text with right FAQ. We extended this framework to match SMS queries with Cross-language FAQs. Results are promising for monolingual retrieval applied on English, Hindi and Malayalam languages.