Abstract: This paper addresses the issue of search of definitions.Specifically, for a given term, we are to find out its definitioncandidates and rank the candidates according to their likelihood of beinggood definitions. This is in contrast to the traditional methods ofeither generating a single combined definition or outputting allretrieved definitions. Definition ranking is essential for tasks. Aspecification for judging the goodness of a definition is given. In thespecification, a definition is categorized into one of the threelevels: good definition, indifferent definition, or baddefinition. Methods of performing definition ranking are alsoproposed in this paper, which formalize the problem as eitherclassification or ordinal regression. We employ SVM (Support VectorMachines) as the classification model and Ranking SVM as the ordinalregression model respectively, and thus they rank definitioncandidates according to their likelihood of being good definitions.Features for constructing the SVM and Ranking SVM models are defined,which represent the characteristics of terms, definition candidate, andtheir relationship. Experimental results indicate that the use of SVMand Ranking SVM can significantly outperform the baseline methods suchas heuristic rules, the conventional information retrieval---Okapi, orSVM regression. This is true when both the answers are paragraphs andthey are sentences. Experimental results also show that SVM orRanking SVM models trained in one domain can be adapted to anotherdomain, indicating that generic models for definition ranking can beconstructed.