Feature Based Opinion Summarization using Transfer Learning. Research Perspective

      Аннотация: Opinion mining is used to improve the decision making of new user in various domains such as product, movie, news media, social networking shares etc. Feature based opinion mining rely only on single domain corpus in most of the existing methodology. Feature based opinion mining in two different domain corpuses is complex. The features and Opinion words are extracted with the help of the Part-of-Speech (PoS) tagging tool. The Inter dependent domain relevance (IDDR) technique use removal of redundant features and pruning of irrelevant features from two different domains with the help of the IDDR score and threshold value. Normally data mining and machine learning use training and test data from same domain and have same feature. But the above concept is not hold in all domains due to the lack of labeled dataset. Here the proposed transfer learning method using Exaggerate Instance weighted K nearest neighbor (EIWKNN) algorithm to transfer the knowledge from camera domain to iPod domain for Opinion classification. The summary of two different domains feature with respect to their opinion is generated.

Ключевые слова: opinion mining, Part-of-Speech tagging tool, algorithm, Nearest Neighbor, Inter dependent, domain relevance, Exaggerate Instance, weighted K