A Khmer named entity recognition method by fusing language characteristics

Aiming at the problem of Khmer named entity recognition, we proposed a method fusing Khmer entity characteristics based on the universal feature templates. For the relatively stable entity that is formed of time expressions and digital expressions, we recognize it using artificial rules; For the complex entity that is formed of names, locations, and organizations, we use Conditional Random Fields algorithm, taking word, part of speech, contextual information and Khmer entity characteristics into consideration, to build a complex entity recognition model to recognize it. Experimental results show that the named entity recognition method fusing Khmer entity characteristics has a better effect.