Dynamic Programming: A Method for Taking Advantage of Technical Terminology in Japanese Documents

摘要

We introduce a new similarity measure based on dynamic programming, intended for technical terms such as machine translation system, which are quite common in technical writing. We compare our proposal with systems which use standard IDF cosine similarity, but on different vocabularies. The dynamic programming method is relatively strong when the query contains a single long technical term, and none of the words in the term are particularly good keywords.