| Semantic relateness is a basic research field of natural language process-ing, one of the key technologies of the text intelligent processing and analysis, which focus on the degree of semantic association between the words in the text. Semantic relateness analysis can effectively improve the traditional text processing and analysis, which used to ignore the semantic association between words in the text. This paper mainly studies the calculation of semantic relate-ness between words based on supported text corpus and its application in the text intelligent processing and mining.We firstly discussed the related technologies of word semantic relateness mining in text, and then,compared the strengths and weaknesses of the existing calculation methods. On this basis, we have accomplished the following three aspects of researches:1. Based on the complex network model named Activation Force (AF) and the co-occurrence of words in the context, a Dynamic Word Semantic Network (DWSN) method is proposed for the analysis of the semantic relateness be-tween words in the specific application environment. Experiments shows that DWSN can greatly improve the precision of the semantic relateness extracted while cost less time in caclulation.2. Base on the DWSN algorithm, a potential relationship entity relation-ship analysis method based on semantic analysis is proposed, to mine the re-lationship named entities implied in the related context. This algorithm has been used in the Campus Object Search Engine (COSE), for the mining of the potential social relations of the teachers.3. Base on the DWSN algorithm, a feature based transfer learning method is proposed to proceed the domain adaption in text classification. We unified the feature spaces of both training and testing corpus, by selecting features with the same semantic meaning. Experiments show that our proposed algorithm can improve the classification accuracy of10%-20%compared to the traditional classification algorithms. |
PDF Full Text Request