Research On Semantic Reconstruction Of Specified Domain Words Based On Hownet And Label Semantics

To solve the problem of semantic reconstruction of specific domain words,this paper constructs a semantic reconstruction algorithm for specific context words based on the semantic representation of words in the HowNet and the study of label semantics.The semantic of domain words is reconstructed by means of label semantics.The specific method is to reconstruct the sememes in the HowNet into the semantic description of the specific domain words,and gives the appropriate degree of each semantic description subset to the specific domain words.At the same time,the semantic similarity calculation method is proposed.Compared with the previous similarity calculation results,the results are more consistent with the understanding of the semantic similarity of words in specific context.The main research is as follows:The fundamental difference between the fuzzy theory of Zadeh and the label semantics of Lawry is studied.Fuzzy theory solves the problem of fuzzy phenomenon which cannot be described by classical set.The membership degree of fuzzy set element to fuzzy concept is proposed by the value of membership function,and we think it is an objective fuzzy theory.Lawry proposed a subjective vague theory in 2004,the suitability of vague concepts is described by the label subset.The appropriate degree is judged by the subjective understanding of participants,and we think that it is a subjective vague theory.A more complex vague concept can be described by label elements with logical operations,which Lawry called method label semantics.In this paper,the semantic algorithm of specific domain words is constructed by means of label semantics.That is to say,in a specific context,sememes are reselected according to the semantic meaning of the words and the suitability of a subset of sememes to the semantic is evaluated.According to the evaluation results,the final set of sememes is given,which is the semantics of the word in the specific domain.The studies of this paper is based on the semantic expression method of the words in HowNet.In the HowNet,the semantics of words are represented by the subset of sememes,which is the smallest unit to describe the semantics.Semantic similarity calculation is a general method to understand the semantics of words.In this paper,the problem of semantic similarity calculation is studied.The distance between sememes is used to calculate the semantic similarity in the HowNet,which is a common calculation method,But it may not be suitable in specific context,because the semantics of a word may change as the environment changes.In this paper,a similarity calculation method based on label semantics is proposed,and the semantic similarity of the words is calculated by the appropriate degree of sememes subset.The results are more consistent with the semantic similarity expressed in specific context.