Novel harmony search-based algorithms for part-of-speech tagging

摘要

As a fast and high-quality tagger algorithm is a crucial task in natural language processing, this paper presents novel language-independent algorithms based on harmony search (HS) optimization method for handling the part-of-speech (PoS) tagging problem. The first proposed algorithm is a framework for applying HS to PoS-tagging which is called HSTAGger. By modifying HS algorithm and proposing more efficient objective functions, two improved versions of the HSTAGger are also introduced. In addition, a novel class of problematic words called erroneous as well as a method of handling them is proposed for the first time to the best of our knowledge. To demonstrate the effectiveness of the proposed algorithms, we have applied them on standard annotated corpus and compare them with other evolutionary-based and classical PoS-tagging approaches. Experimental results indicate that the proposed algorithms outperform the other taggers previously presented in the literature in terms of average precision.