Research On Ontology-based Mathematical Expression Retrieval Technologies

In many fields such as scientific research, education and engineering, mathematicexpression represents complex and strong semantic information.When to be retrieved,human have expectation to recognize the two-dimensional structure and the semanticscontent. However, most existing mathematical retrieval system are based oncharacterized expression and context-displaying, which is difficult to meetunderstanding for mathematical expression. Thus, it has practical significance toachieve retrieval on expression semantics for learning and work.Based on the mathematical expression ontology library which supports semanticcontent and incidence relationships, the searching conditions are expanded in this thesis,hoping to get complete retrieval on mathematical expression. The major research worksin this thesis are summarized as follow:1) By analyzing the processes of retrieval, a retrieval model based on ontology isproposed in this thesis. The major idea of the model is that summarize four incidencerelationships for the multi-information of mathematical expression and constructontology library based on an improved wildcard character. Based on the library, thesemantic conditions are expanded and the key vocabularies are returned to users to getaccurate operation or transfer to retrieval frame directly in this research, achievingsemantics retrieval for mathematical expressions.2) A matching method of DOM-based mathematical expression is revised andimproved in this thesis for achieving the semantic extension of the above model. Themajor idea is that operational cost and corresponding fitness are obtained according tothe differences among edited nodes and parent nodes, and then calculating the totalvalue of the operating costs and the similarity, completing the accurate expandingpurpose for semantic expression. Meanwhile, the time for matching can be reduced andthe efficiency can be enhanced by using the incidence relationship and up–downrelationship in the ontology library.3) Based on the primary works mentioned above, Protege4.3is used to constructan initial mathematical ontology library in this thesis, and do experiment based onLucene Retrieval frame, verifying the reliability and efficiency of the mathematicalsemantic retrieval model and methods based on ontology. The contrast experiments arecarried out for the improved matching method of DOM-based. The experimental data show that the improved method has obvious improvements from responding time, recallration, precision ration and F-measure.