Study On Term Knowledge Base Construction Based On Semantic Analysis

A knowledge base is a collection of organically organized knowledge.No matter how the years change and what kind of computing model dominates,it is an integral part of the development of artificial intelligence.A semantic knowledge base is the basic resource of natural language processing tasks.It contains machine-readable semantic knowledge.It not only provides rich formal background knowledge for tasks such as semantic computing and semantic reasoning,but also provides strong support for natural language understanding.Traditional large-scale semantic knowledge bases are mostly constructed manually and belong to the general domain.With the rapid development of artificial intelligence,the demand for knowledge is now more extensive,and higher demands are placed on existing technologies and resources.That has led to the gradual expansion of the knowledge base from the general field to the specialized field and the gradual transformation from inefficient manual construction to highly efficient automatic construction in order to meet the new challenges brought by the technological revolution.Therefore,it is an urgent and meaningful task to study a method of efficiently constructing a domain semantic knowledge base.This paper presents a method of assistantly constructing a semantic knowledge base of aviation terminology based on the semantic theory system of HowNet.The method realizes the task of constructing the term semantic knowledge base from the semantic level of the term according to the bottom-up construction idea.It greatly improves the construction efficiency.The main work of this paper includes the following aspects:According to the characteristics of terms,the knowledge base construction process can be divided into four key processes,including the analysis of the term dependency structure,the assistant selection of the meaning of words within the term,the assistant judgment of the relationship between two words within the term and the generation of the term DEF(the definition of concept).The first three processes complete semantic analysis of the term.Aiming at the problem of polysemy of terminology,this paper proposes a method of semantic disambiguation based on the collocation of words.We combine the method into the process of selecting each word meanings within the term.Experimental results verify the validity of this method.In addition,this paper proposes an assistant judgment method of the dynamic role.It is based on the combination of a maximum entropy classifier and a method based on similarity.This method is created from two different levels that include the statistics level and the semantic level.Its aim is to determine the relationship types for each relevant unit within the term and fill up the internal relationship in the triples.Experimental results show the effectiveness of the method.For the process of term DEF generation,this paper proposes a term DEF generation method based on dependency tree.The method maps all the relational triples in the term to the term DEFs in accordance with the provisions of the KDMLof HowNet based on the internal dependency structure of the term.We also implemented the term semantic knowledge base assistant construction system and constructed 2,000 term DEFs with this system.The effectiveness of this method is validated by computing the similarity between two terms that are random sampled.