Part-of-speech tagging for web search queries using a large-scale web corpus
Atsushi Keyaki, Jun Miyazaki · 2017
This paper proposes an accurate part-of-speech (POS) tagging method for Web search queries using the sentence level morphological analysis results of a large-scale Web corpus. POS tagging is a fundamental technique for analyzing queries; however, the existing NLP tools often fail to correctly identify POS tags because queries are not based on natural language grammar. We propose a method not affected by the queries' characteristics lacking capitalization and free word order with the term-POS database (TPDB). Experimental results show that the proposed method significantly outperforms those using existing NLP tools and the state-of-the-art method. In addition, the data set we created is expected to be useful for future researches on both POS tagging systems to queries and IR systems leveraging POS tags.