Research And Application Of Domain Terminology Semantic Knowledge Base Construction Method

The semantic knowledge base provides resources that computers need to understand text,provides a strong support for semantic analysis and computing,and it is widely used in the fields of word sense disambiguation,information retrieval,machine translation and the others.At present,most of the existing large-scale semantic knowledge bases belong to common sense knowledge bases,which cannot meet the needs of professional knowledge in the rapidly developing artificial intelligence and other application fields.The internal Event Role labeling of the term is the key link to construct the domain term semantic knowledge base,the quality of the labeling results directly affects the scale and quality of the knowledge base.How Net is the most mature semantic knowledge base for computer processing in the field of Chinese.Therefore,based on the existing domain term semantic knowledge base construction method and How Net theory system,this paper mainly researches the internal Event Role labeling method of domain term.On the issue there are too many varities of Event Role to label,a method for labeling based on KNN(K-Nearest Neighbor)is proposed.This method performs sample pre-selection by analyzing the DEF(the definition of concept),and combines DEF-based semantic similarity and word vector-based context distribution similarity in the nearest neighbor sample selection stage.The experimental results show that the accuracy rates P of 1-Best,3-Best,and7-Best are 67.57%,86.00%,and 94.17%,respectively,and the average reciprocal ranking MRR is 0.7764,which is better than existing research results.On the issue that higher labeling costs,this paper proposes a method for labeling based on active learning,selects potentially valuable samples for labeling.This method takes the KNN classification algorithm proposed in this paper as the benchmark classifier,the uncertainty measure as the sampling basis,the representativeness of the sample based on the DEF similarity calculation measure is fused.The experimental results show that the number of samples to be labeled is reduced by 62.03% when the same accuracy is reached,which can effectively reduce the time and cost of manual labeling.Finally,this paper designs and implements a domain term semantic knowledge base construction system,and applies the system to the field of aviation,and the semantic knowledge base is analyzed using similarity calculation,proving that the validity of the method in this paper.