计算机应用与软件Issue(9)：32-34,56,4.DOI:10.3969/j.issn.1000-386x.2013.09.010
基于 Co-training训练 CRF模型的评价对象识别
OPINION TARGET RECOGNITION METHOD BASED ON TRAINING CRF MODEL BY CO-TRAINING
摘要
Abstract
Opinion target refers to the object or object ’ s property modified by the evaluation words in certain commentary .In order to recognise opinion targets , we present a Co-training based CRF model training method .First the method manually labels a few original dataset , then uses the Co-training means to automatically recognise the unlabelled data so as to extend the labelled training data .The CRF model is trained through original labelled dataset and Co-training labelled dataset .In car domain, the precision rate and recall rate of opinion target recognition achieve 67.483%and 67.832%respectively in corpus of car commentaries to be labelled .关键词
CRF模型/评价对象/特征模板/Co-trainingKey words
CRF model/Opinion target/Feature template/Co-training分类
信息技术与安全科学引用本文复制引用
张彩琴,王素格,乔磊..基于 Co-training训练 CRF模型的评价对象识别[J].计算机应用与软件,2013,(9):32-34,56,4.基金项目
国家自然科学基金项目（61175067，60970014）；山西省自然科学基金项目（201001102-1）；山西省科技攻关项目（20110321027-02）。 （）
