基于双代价参数SVM的生物医学文本指代消解研究

生物医学文本中的指代消解是生物医学信息抽取领域的一个重要组成部分.通过引入双代价参数对基本SVM方法进行改进,并在FlyBase语料集上进行了测试,准确率、召回率、F值分别达到53.9%、69.5%、60.7%.同时研究了特征向量的选择和取值对于实验结果的影响.最后与其他先进方法进行了对比.结果表明,在同样的语料上,基于双代价参数SVM方法优于其他先进的方法.

Research on anaphora resolution in biomedical texts based on doubly-cost parameter SVM

The anaphora resolution in biomedical text is an important part of biomedical information extraction. The basic SVM method was improved by introducing doubly-cost parameters, and tests on the FlyBase corpus were conducted. The precision, recall and F -value are 53.9% 69.5% and 60.7% respectively. In the meantime, the influence on the final results of different selections of the feature vectors are studied. Finally compared with other state-of-the-art methods on the same corpus it is shown that the proposed SVM method based on doubly-cost parameters is superior to the others.