Chinese Standard Comparative Sentence Recognition and Extraction Research

摘要

Information extraction is the first and foremost important task of Standard Knowledge Mining. The paper focuses on comparative sentence recognition and extraction. There are three steps, respectively Comparative Sentence Recognition, Technical Index Parameter Recognition and Technical Index Extraction. At first, we search the standard by keywords from a feature set lists in order to categorize the documents into its specific class. In addition, we build the regular expression to spot technical index parameter. Lastly, we treat the technical index extraction as a sequence labelling problem and treat the keyword, noun phrases, and theirs position as features training by CRF model. The final experiments show that the result performs very well in standard document comparative sentence recognition and extraction.