一種適應(yīng)短文本的相關(guān)測(cè)度及其應(yīng)用:針對(duì)博客社區(qū)和BBS 論壇充斥Web 垃圾信息的問(wèn)題，提出相關(guān)度向量空間模型cVSM，并以此作為評(píng)論的特征，采用支持向量
機(jī)分類(lèi)算法自動(dòng)識(shí)別垃圾評(píng)論。cVSM 包括一種適合短文本的相關(guān)測(cè)度，用于衡量評(píng)論和文章的語(yǔ)義相關(guān)程度。在中文博客測(cè)試集和中文
BBS 測(cè)試集上的實(shí)驗(yàn)結(jié)果表明，相比純粹使用評(píng)論文本特征的方法，應(yīng)用該模型時(shí)F1 至少提高6%。
關(guān)鍵詞：博客；垃圾評(píng)論；支持向量機(jī)；文本挖掘；相關(guān)測(cè)度
Relevancy Coefficient and Its Application Adapted to Short Texts
HE Hai-jiang
(Computer Center, Changsha University, Changsha 410003)
【Abstract】A relevancy coefficient vectort space model named cVSM is proposed to aim at Web spams which flood in blogosphere and forums.
The cVSM whose components are employed as features of comments and the support vector machine classification algorithms are used to
automatically identify comment spams. The relevancy coefficient included in the cVSM is presented, which is used to evaluate relevancy grade of
posts and comments. Chinese blog dataset and Chinese BBS dataset are tested. Experimental results show that compared with traditional method the
F1 has been improved at least 6% by this way.
【Key words】blog; comment spam; support vector machine; text mining; relevancy coefficient