Based On The Management Of Multiple Semantic Dictionary And Implementation Method

In the field of Natural Language Processing, dictionary management model has play a decisive role position, whether the information retrieval, machine learning,semantic analyzing system, semantic dictionary is one of the main sources of data information processing. However, the creation and maintenance of a dictionary is a basic and important work, it is used as the underlying data sources, is the foundation of Natural Language Processing processing. Therefore, how to make the optimal dictionary management in quality and performance, it is very important.At present, most of the dictionary management system mostly HowNet hyponymy, synonymy and other 16 kinds of relationship based on the morphological basis, it mostly is the reflection of the relation between word and word, concept of relationship model can’t really from the angle of semantics of words to construct words, such as from the semantic point of view to represent the concept of words, has been the focus of many experts and scholars to study. Based on the previous research,the dictionary management system model for more in-depth research, put forward a abstract concept parent class and subclass description of abstract concept concept relationship model, the real from the semantic point of view to represent words links among concepts, the main work is as follow:First of all, the paper introduces the research status of the dictionary management system at home and abroad. The analysis of all kinds of semantic relations between the words in Chinese sentences express, and points out the inadequacies of the relationship of words representation.Secondly, presents a descriptive words between the concept model from semantic collation abstraction relation model, and gives the abstract concept tree semi automatic expansion methods, and the abstract concept classification test method.In view of the existing dictionary management model can’t really from the perspective of semantic words describe the relationship between the concept of using words, the abstract concept model can be put in the same industry words divided into abstract concepts and abstract concept of parent class subclass relationships, and expressed in the abstract concept tree, in order to solve the common dictionary management model can’t be criticized in a timely manner update the dictionary system, this paper gives the abstract concept relationship tree semi automaticexpansion method. At the same time as the dictionary maintenance is mainly through the manual, there will inevitably be a mistake, will return to the words does not match the category of words in the collection. This will be in the late word tagging, semantic analysis have been affected work. In view of this phenomenon, this paper gives an abstract concept classification method of calibration. And classification of the calibration method of abstract concepts were tested and analyzed, and the comparison and other methods, the experimental results show that the modified calibration method of abstract concept classification.Finally, According to the abstract concept of relation model, design a dictionary management system.The dictionary management system is to construct an abstract concept based on the relationship model, each industry under the words according to the abstract concept of the parent class and subclass of abstract concept description, this word in dictionary is not isolated, but the formation of a certain "network".