Phrase semantic mining method driven by directed graph meaning guiding model

Abstract

According to the phrase semantic mining method driven by the directed graph meaning guiding model, logic structure representation of a typical ontology language Word-Net is achieved through a Sem-Graph data model, and on this basis, modeling work of the Word-Net ontology language is achieved. The method comprises the following steps: performing semantic-level phrase structure mining on natural language text data based on a Sem-Graph model; modeling statements in a natural language by using a semantic-oriented semantic model, realizing statement-level semantic graph data structure description;defining a phrase semantic structure on a semantic graph; and realizing mining of frequent phrase semantics by using a frequent sub-graph mining algorithm. According to the method and the system, thedocuments can be properly described and accurately summarized, so that each minimum processing unit has independent and relatively complete semantic features, high-quality phrases related to the fieldcan be mined from a large amount of text data, and the increasing phrase semantic mining requirements are fully met.

Description

technical field [0001] The invention relates to a phrase semantic mining method, in particular to a phrase semantic mining method driven by a directed graph meaning-oriented model, and belongs to the technical field of phrase semantic mining. Background technique [0002] With the rapid rise of the new generation of mobile Internet technology, more and more people like to share some interesting and important news through social platforms, or express their views on some high-profile and important social events. According to statistics, Weibo generates hundreds of millions of texts every day, and text data can be decomposed into a series of related fragments, which are not only concise in form, but also contain rich and valuable information. Among such a large number of texts, some texts have clear descriptions of events or viewpoints, while others are very difficult to understand. In order to properly describe and accurately summarize these text data for big data analysis or ...

AI Technical Summary

Problems solved by technology

[0006] On the whole, the classification of information categories in the prior art has the following defects: First, the text data mining methods in the prior art cannot properly describe and accurately summarize documents, and cannot make each smallest processing unit have an independent and relatively Complete semantic features cannot mine domain-related high-quality phrases from a large amount of text data, and cannot meet the growing demand for phrase semantic mining; second, the existing phrase mining work mainly focuses on the mining of specific phrases in natural language , Lack of understanding the phrase structure in natural language from the perspective of semantics, did not introduce a semantic-oriented directed graph structure data model, cannot re-understand phrases in natural language from the perspective of semantics, cannot recognize natural language from the perspective of semantics, and cannot understand sentences Define its corresponding semantic graph data structure; third, the existing technology cannot perform semantic-level phrase structure mining on natural language text data, and does not use a semantic-oriented semantic model to complete the modeling of natural language sentences, and cannot realize sentence-level semantics The graph data structure description cannot define the semantic structure of phrases on the semantic graph, and the frequent subgraph mining algorithm cannot be used to mine the semantics of frequent phrases; the fourth is that the existing technology does not affect the integrity of semantic information. The simplification of the data structure cannot improve mining efficiency, does not have the characteristics of modularization, does not have universal applicability, and the method steps and calculations of the existing technology are relatively complicated

[0007] Phrase mining work in the prior art mainly focuses on the mining of specific phrases in natural language, and lacks understanding of the phrase structure in natural language from a semantic perspective

Images