【摘要】 第六届中文倾向性评测(COAE2014)新增了"面向新闻的情感关键句抽取与判定"任务,为了保证评测数据的真实有效性和评测任务的合理性,本文采用了随机抽样法和交叉验证法进行数据采集和标注,制定了符合任务特点的评测指标,提出了基于关键词累积得分和朴素贝叶斯分类器的baseline方法。本评测任务吸引了国内13家研究机构和高校的共同参与,提交的22个结果中有一半的成绩超过了baseline,但是最好的结果离人工标注结果还有很大差距。更多还原

【Abstract】 Task " News Sentiment Sentence Analysis Evaluation" was added in the sixth Chinese Opinion Analysis Evaluation(COAE2014).In order to ensure that evaluation data real and effective and the task reasonable,this paper used random sampling method and the cross validation method for data collection and annotation,and established evaluation indicators accord with the characteristics of the task.The baseline method based on keywords cumulative scores and Naive Bayesian classifier is proposed.This evaluation task had attracted domestic 13 research institutions and universities to participate,who submitted 22 results and half performed better than the baseline,but the best result was still far from the marked results.更多还原