中文观点挖掘中主观性关系抽取.pdfVIP

22 2 V ol . 22, No . 2 2008 3 J OU RNA L OF CH INESE IN FO RM AT ION PR OCESSIN G M ar., 200 8 : 200 8) , , , , 200433) : 本文所针对的具体任务是抽取评价词和目标对象之间的关联关系所采用的方法是将同 句子中共现的 评价词与评价对象作为候选集合, 应用最大熵模型并结合词词性语义和位置等特征进行关系抽取我们将关系 抽取 引入观点挖掘, 所提出的方法 定程度上解决了指代消解以及评价对象遗漏的问题实验结果表明该方法的 F 值比取最近评价对象的 Baseline 方法有了15% 的提高, 并且发现程度副词能够帮助提高主观性关系抽取的性 能 : 计算机应用; 中文信息处理; 观点挖掘; 关系抽取; 最大熵 : T P39 1 : A Subjective Relation Extraction in Chinese Opinion Mining ZHA NG J ianfeng, ZH AN G Q i, W U Lide, H U AN G Xuanjing Depa rtment of Comput er Science, Fudan U niver sity, Shang hai 200433, China) Abstract: T his paper presents a nov el met ho d to ext ract t he subjectiv e r elationship betw een opinio nbearing t erm s and opinion tar gets . T his metho d ex tr act ed the pairs o f opinio nbearing ter m s and opinion tar gets as the candidate set , and then employed the max imum ent ropy mo del t o combine lex ical, par t of speech , semantic and po sitio nal f eat ur es derived f rom tex t . Our method incor po rates relation ex traction into o pinion mining and solves the pro blem of cor ef er ence and omitt ing o f opinion tar gets to so me extent . T he ex periments sho w ed that the F v alue of o ur met ho d is 15% hig her than that of Baseline w hich t akes the nearest opinion ta rg et as the r eal t arg et , Besides, the ex periments found that the intensif iers can improv e the perfo rmance of subjective relation ex tractio n . Key words: computer application ; Chinese infor mat ion pr ocessing ; o pining mining; r elation ex tr act ion ; max imum entr opy