Research And Implementation Of Chinese Open Relation Extraction

Information extraction, natural language understanding, information retrieval and other applications need a better understanding to the semantic relations between entities. Relation extraction is an arduous and very important task in information extraction task. It is also a hot topic in information retrieval and natural language processing. It has very important theoretical significance and wide application value.With the emergence of large-scale web corpus,the research on relation extraction has evolved from Traditional relation extraction (TRE) to open relation extraction(ORE).The emergence of ORE has attracted the attention of many scholars at home and abroad, and made many research results. The research on the open relation extraction of English corpus is very mature. However, due to the complexity of Chinese language and the difference of Chinese and English language, the performance of open relation extraction for Chinese corpus is relatively low. Based on the analysis and comparison of current algorithms for ORE, this thesis presents a new algorithm for ORE. Specifically, the work of this thesis mainly includes the following aspects:1.This thesis analyzed and compared various algorithms for ORE. According to the difference relation extraction methods, this thesis summarized the existing methods for ORE into four different types: the method based on pattern matching,the method based on machine learning, the method based on inference and the method based on sentence simplification. This thesis conducted in-depth analysis and comparison of four different methods for ORE.2.This thesis presented a relation extraction algorithm based on the fusion of rule and syntactic dependency. In order to realize the effective nary relation of Chinese corpus, this thesis explored how to extract the nary relation based on the combination of rules and syntactic dependency according the Chinese Annotation Guidelines for Relation Detection Automated Content Extraction (ACE) and also the Characterization of Chinese word, sentence pattern and syntactic constituent of Chinese sentence.3.This thesis designed and implemented a Chinese relation extraction system.Based on the fusion of rule and syntactic dependency, this thesis implemented the ORE system. Besides, this thesis designed and implemented a baseline system based on training corpus and pattern matching method. Multiple experimental results show that the performance of our proposed system is improved significantly when comparing with ZORE system and baseline system, and the validity of the relation extraction algorithm proposed in this paper is also proved.