So many data cleaning techniques have been used for cleaning data in the database. In approaches such as dynamic programming, information is matched by literally comparing and scoring pairwise string symbols in the data records. Higher scoring pairs of strings are matches, while low scoring pairs are mismatches. This paper presents an effective technique in which the word with similar meaning, misspelled, wrong spelling will be replaced with the correct word as per the requirement. For this matching, latent semantic indexing approach is used using singular value decomposition technique.