Font Size: a A A

Term Recomendation And Visualization Based On Semantic Similarity Computation

Posted on:2018-08-27Degree:MasterType:Thesis
Country:ChinaCandidate:T WeiFull Text:PDF
GTID:2348330536458090Subject:Software engineering
Abstract/Summary:PDF Full Text Request
With the rapid development of the information age,the world becomes mobile and intelligent.The network information data are exploding,and the terminology is constantly updated and development.The richness of information culture promotes the diversity of the meaning of terms.The growth of the scale of the term data has certain influence on the formulation of the ISO standard of international standard organization.The term expert needs a term exchange platform to realize the update of the term and the formulation of the work.The method of semantic similarity calculation of the term has a fundamental role in the development of other disciplines,such as information retrieval,Machine Translation,artificial intelligence and so on.At present,the method of semantic similarity calculation is mainly based on the organization of data,structured data and unstructured data.The structured organization of data includes ontology,Hownet,WordNet,etc.The organization form of unstructured data,often large-scale data,no fixed structure,the calculation method of the mainstream semantic similarity is used to train the model through machine learning,and then call the semantic similarity calculation model for term.In this paper,the research focus on the semantic similarity of term for the structured data based on the ontology and the large scale unstructured data.The mainly content as follow:(1)The method of semantic similarity calculation based on structured data could not consider all kinds of facts as result of accusing the problem of semantic similarity calculation inaccurate.Such as,the semantic similarity calculation based on the features of word formation,the semantic similarity calculation based on the features of syntax,the semantic similarity calculation based on the context.In this paper,it proposed the method,which is improved the hybrid semantic similarity algorithm based on ontology.This method calculates the weight value through drawing on the idea of fuzzy optimization so that the accuracy was increased.Meanwhile the method of this paper proposed will be applied on the term recommendation.Before the term will be recommended by expert,they need to calculate the semantic similarity of term in order to determine whether in standard documents of terms are synonyms or near synonym.Then decide whether it is necessary to update the term file.(2)With the advent of the era of big data,the semantic similarity calculation method of large-scale unstructured data corpus has gradually become a research hotspot.There is another important research spot that extract the word for semantic similarity of term from the large-scale data and visual those words.In this paper,the method of semantic similarity based on vector will be used for the large-scale unstructured data.Training the model based on the Word2 vec.The text is represented by a vector of words in the corpus.The semantic similarity calculation will be based on the vector.Through this method,it will obtain the word for the semantic similarity of term.In the experiment,the method of this paper could achieve the expected goal.When using the method obtain the semantic similarity word,its relationship network will be visualization by calling Prefuse components.In this way,it is convenient for the term workers to explore the potential relationship between the terms,and also lay the foundation for the later knowledge mapping.
Keywords/Search Tags:Term, Semantic Similarity, Structured, Unstructured, Visualization, Term Recommendation, Knowledge Mapping
PDF Full Text Request
Related items