Word Sense Disambiguation-based Sentence Similarity Chukfong Ho, Masrah Azrifah Azmi Murad Department of Information System University Putra Malaysia Rabiah Abdul Kadir, Shyamala C. Doraisamy Department of Multimedia University Putra Malaysia Coling 2010
22
Embed
Word Sense Disambiguation-based Sentence Similarity
Word Sense Disambiguation-based Sentence Similarity. Chukfong Ho, Masrah Azrifah Azmi Murad Department of Information System University Putra Malaysia. Rabiah Abdul Kadir , Shyamala C. Doraisamy Department of Multimedia University Putra Malaysia. Coling 2010. Outline. Abstract - PowerPoint PPT Presentation
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Word Sense Disambiguation-based Sentence Similarity
Chukfong Ho, Masrah AzrifahAzmi Murad
Department of Information SystemUniversity Putra Malaysia
Rabiah Abdul Kadir, ShyamalaC. Doraisamy
Department of MultimediaUniversity Putra Malaysia
Coling 2010
Outline• Abstract• Introduction• Related Work• Sentence Similarity• Experimental Design• Results and Discussion• Conclusion
Semantic Text Similarity (STS)model (Islam and Inkpen, 2008).
作者準備的 model 由上述修改而成
Sentence Similarityfor string similarity measure
• 計算對象 :two words ( 最高 =1) • a , b : 分別是兩句話長度 ( 刪除 stop words 後 )• l (x):x 的長度• i , j : 字串中字元位址• LCS: 最長相同子字串• (Islam, Aminul, and Diana Inkpen. 2008.)
Sentence Similarityfor string similarity measure
Sentence Similarityfor Adopted word similarity measure
• l :wi 和 wj 之間在 Wordnet 中語義距離。• t : 表示兩個 sense 關係 ( 數字 ) ( 上位語 (hypernyms)/ 下位語
(hyponym)/ 同位語 (synonym)/x 包含在 y 的語義關係 (holonym)(x和 y 為英文字串 )) 。
• 透過經驗法則設定 :α=0.9,β=0.85,γ=12 。• Based on Wordnet2.1 • 此 model:YP(Yang and Power.2005)• 作著修改 YP 部分為 MYP( 改變其 length ) 。
word similarity measure 問題
• 舉例 dog 和 cat 作比較 :– 句 1:The dog barked all night.– 句 2:What a cat she is !第一句的 dog:”a member of genus Canis that has been Meanwhile by man since prehistoric times”第二句的 cat:”a spiteful woman’s gossip”dog 和 cat 在這兩句真正語義距離為 7 ,但一般最短路徑狀態下距離為 4 ,造成頗大的差距。
Sentence Similarityfor WSD-STS
• Baseline:STS
• a,b : 個別為 a 和 b 語句長度。• δ : a , b 之間出現相同的字數。• : 前 : 最高的 match score 後 : 取兩 sense 距離較近