| Math Information Retrieval(MIR)is an important research content of applying information retrieval technology to the field of scientific and technological information with mathematical expressions as an all-purpose language.Due to the two-dimensional structural characteristics of mathematical expressions,the high-performance and practical MIR system remains to be further researched and developed.Aiming at the problems existed in the most MIRs that their matching operations of mathematical query expressions and candidate document sets containing mathematical expression are majorly based on mathematical expressions which results in the mathematical semantics contained in the context of mathematical expressions be ignored and the retrieval performance be affected,a fusion retrieval method that combines mathematical expressions and contextual text was proposed in this thesis.Firstly,the Chinese scientific and technical documents containing mathematical expressions were traversed,and the mathematical texts in the documents are extracted via the information extraction method based on rules and statistics.Secondly,these mathematical texts were converted into mathematical expressions in LaTeX format by constructing a mathematical transformation dictionary.And these expressions were furtherly transformed from mathematical texts into binary tree structures together with the original mathematical expressions in the document for extracting the retrieval features and building index.Finally,the corresponding matching algorithm is designed to realize the fusion retrieval of mathematical expressions and contextual text.Through the experimental analysis to the methods of searching only with mathematical expressions by employing 8,532 Chinese scientific and technical documents,the evaluation value of the retrieval results of the proposed method is improved,which indicates that the method is effective for improving the performance of the mathematical expression retrieval system. |
PDF Full Text Request