Engineering in genomics: text comparison based on dynamic programming

Inspired by BLAST and related sequence comparison algorithms, we have developed a method for the direct comparison of query text against database text as an improvement upon traditional keyword-based searches. The primary application of our implementation, eTBLAST, is to better select those database entries (abstracts, in the case of MEDLlNE) of most relevance to a given query. eTBLAST takes as input natural text instead of keywords, allows refinement of retrieved hits through iteration, can be applied to any text (demonstrated here on biomedical databases), and allows inspection of the local space around a query through simple visualization methods.