Research On Comparative Sentences Ellipsis Identification And Element Extraction

With the rise of web2.0, forums, microblog and paste not only makes people become an information share, also let people to become an information provider. Huge amounts of comment text appears in the network, Comparative sentences is a common sentence pattern, which often used to contrast between two or more things, to express people’s emotional tendency to different things. Comparison element extraction can provide data resources for business to market similar products of comparative analysis, and provide decision support for ordinary consumers to purchase goods. When people write reviews, pursuit of simplicity, often using ellipsis, due to compare elements extraction performance is not high for the computer, In order to solve this problem, this paper mainly discusses the elements of comparison ellipsis identification and extraction for comparative sentences in-depth study. The major work of this thesis includes:(1) Comparison Element Ellipsis Identification Based on Rules and Sequence PatternsAccording to the characteristics of comparative sentence patterns, Summarized the comparison of words set. On this basis, Constructed quintuple identification rules and sequence pattern mining algorithms. Moreover, making full use of the advantages of two methods, a mixed strategy is designed based on quintuple method and sequence pattern. Comparative experiments on the COAE2013 corpora, the experimental results indicate that the mixed strategy is better than sequential patterns and quintuple method for comparison element ellipsis identification.(2) Comparison Element Extraction Based on the Comparison Element EllipsisBased on the investigation for the comparative sentences elements ellipsis, summarizes two forms of ellipsis. One is the explicit ellipsis for the simplicity of expression, the other is comparative sentences attribute of implicit ellipsis. For explicit ellipsis mainly through selecting different sizes of window case sentences tagged objects and attribute as the corresponding comparison element, respectively. The second kinds of attribute implicit ellipsis need to speculate from context, mainly through using synonyms Lin implicit properties of expanded library for recovery, and then extraction recovery attribute.(3) Element Extraction System Based on the Restricted Domain Knowledge of ComparisonThe Chinese comparative sentences identification method based on sequential patterns, and the methods of comparative sentences ellipsis identification and recovery, developed a set of comparative sentences identification, comparative elements ellipsis identification, comparative elements extraction a factor extraction as one of the comprehensive research system. Can be used provide technical support for product analysis.