Phonetic matching and syntactic tree similarity based QA system for SMS queries

Currently there have been many QA systems for SMS queries with large archive but none is addressing the most commonly present phonetic noise in SMS queries. However, Finding similar questions in the QA archive is not trivial especially in the presence of phonetic noise. This paper proposes a solution to handle the noise including phonetic noise. We present a technique to handle semantic variation by developing new phonetic algorithm that uses Soundex and Metaphone algorithm; In addition we have modified Longest Common Subsequence problem to raise the level of similarity between noisy words of SMS and corresponding dictionary words. Following this approach it also handles syntactic variation in question formulation without any SMS normalization.