HITSCIRRunCOAE2015微博观点句识别任务分析系统.PDFVIP

HITSCIR_Run ：COAE2015 微博观点句识别任务分析系统* 1 2 1 1 李泽魁 ，赵妍妍 ，秦兵 ，刘挺 1 哈尔滨工业大学计算机科学与技术学院信息检索研究中心，哈尔滨，150001 2 哈尔滨工业大学机电学院媒体系，哈尔滨，150001 E-mail: {zkli, yyzhao, qinb, tliu}@ir.hit.edu.cn 摘 要：文本倾向性分析作为自然语言处理领域研究的热点问题之一，目的是通过文本来分析人们的情感 倾向。本文主要研究第七届中文倾向性分析评测 （COAE 2015 ）中的微博观点句识别任务（任务 2 ），通 过词典资源收集、词向量训练、特征抽取模板设计、训练语料构建、分类模型调优与集成等步骤，完成了 一套有监督学习的基于特征抽取的情感倾向性分析系统。本系统在评测任务中取得了较好的成绩。 关键词：情感倾向性分析；特征抽取；情感专属词向量；模型集成 HITSCIR_Run:Sentiment Analysis System of Micro-blog Opinion Recognition Task in COAE2015 1 2 1 1 Zekui Li , Yanyan Zhao , Bing Qin , Ting Liu 1Research Center for Social Computing and Information Retrieval of Computer Science and Technology School, Harbin Institute of Technology, Harbin 150001 2 Department of Media Technology and Art, Harbin Institute of Technology, Harbin 150001 E-mail: {zkli, yyzhao, qinb, tliu}@ir.hit.edu.cn Abstract: Text sentiment orientation analysis is a hot research feild in natural language processing. It aims to annalyze people’s subjective opinions through plain text. This paper mainly focuses on the seventh Chinese Opinion Analysis Evaluation (COAE 2015) in micro-blog opinion recognition task (Task 2). We finally build a sentiment analysis system based on feature extraction algorithm, which is a kind of supervised learning method. More in detail, we did lots of works in training corpus and lexicon resources building, word embedding training, template of feature extraction design, model optimization and ensemble and so on. Our system achieves good results in the evaluation task. Keywords: sentiment orientation analysis; feature extraction; sentiment