這篇論文出自CIKM’06，目的是做movie review mining和summmarization，主要是藉由產生正面與負面的feature-opinion pairs做為此電影資訊的摘要。此篇論文提出的方法是一開始先利用WordNet、IMDB的movie casts和labeled training data去產生feature跟opinion的keyword list，利用此list找出句子中的feature words和opinion words且決定feature words是屬於哪個feature class和opinion words是正面的還負面的；訓練階段會利用pattern mining 找出dependency grammar graph中feature到opinion pair之間的常見pattern，做為測試階段feature-opinion pair找尋的依據。
實驗部分跟Hu and Liu的KDD'04的opinion extraction做法比較，以擷取feature-opinion pair 的precision 和recall做為評估原則。
連結: Proceedings of the 15th ACM international conference on knowledge management CIKM'06
skip to main |
skip to sidebar
研究相關鏈結
依標籤分類文章
- information extraction (8)
- document classification (7)
- 巫孟倫 (7)
- aboutlab (6)
- JhihMing (5)
- association rule (5)
- social network (5)
- GPS (3)
- keyword selection (3)
- text summarization (3)
- Don (2)
- Path Detection (2)
- cfp (2)
- document clustering (2)
- mobile computing (2)
- opinion extraction (2)
- recommendation (2)
- web data extraction (2)
- AI (1)
- DBN (1)
- Data integration (1)
- Schema matching (1)
- Web application development (1)
- coclustering document (1)
- google (1)
- ontology (1)
- query translation (1)
- schama (1)
- semantic web (1)
- web UI (1)
- web modeling (1)
