Research On Product Named Entity Recognition

Named Entity(NE) recognition is an important foundation of Natural Language Processing(NLP) technology. In texts,product models are often confused with alphabet-words and normal numbers etc. So Product Named Entities(PNE) containing product models cause some negative effects on language information processing. This study creates a product named entity recognition module in an existing word segmenting and POS tagging system. Based on the achievements of previous studies and corpus analysis, the author elaborates the definition, composition and variations of product named entity, and gives a detailed feature analysis for product model, which is the most typical part of product named entity. Based on this, regard product models as the core, a strategy of product named entity recognition using a combination of rule-based and statistic-based methods is formed. Meanwhile, the way of integrating the recognition module into the original system is also taken into consideration. In practice, the author tries to use chapter echoing discipline to recognize variations of some product named entities with relatively weak context characteristics. This produces a significant effect. Through experiment, the recognition module made a relatively good result in home appliance, mobile phone and computer digital product areas. Through analyzing the experimental results, some problems to be studied further are also found.