COAE2014情感关键句评测任务和评测数据设计  

Task and Data Designing of Sentiment Sentence Analysis Evaluation in COAE2014

作　　者：杨静[1] 徐蔚然[1] 谭松波[2] 

机构地区：[1]北京邮电大学信息与通信工程学院,北京100876 [2]中国科学院计算技术研究所,北京100190

出　　处：《山西大学学报（自然科学版）》2015年第1期16-23,共8页Journal of Shanxi University(Natural Science Edition)

基　　金：111项目(B08004);国家科技重大专项(2011ZX03002-005-01);国家自然科学基金(61273217);国家博士点基金(20130005110004);北京邮电大学青创基金(2011RC0122)

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。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.

关 键 词：COAE2014 新闻 情感关键句 评测 任务设计 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]