Opinion Mining from Amharic Blog
No Thumbnail Available
Date
2013-04
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Before the Web, organizations have been conducting a survey and people have been asking their family, friends and neighbors for the information (factual or opinion) in order to make a wise decision. With the growing popularity of sophisticated and advanced technologies like the Web, the world has become a single village. It is possible to get documents expressing opinions that are generated, propagated, exchanged, stored and accessed through the Internet. The accumulation of vast and unstructured opinions on the web has been making information acquisition difficult. Opinion mining is the preliminary technique towards tackling this obstacle. It can be performed in one of the three different levels: sentence, document or feature level. Among the three levels, feature level opinion mining is the detail and complex but has a better advantage to meet customers and organizations need. Although there are many feature level opinion mining models that have been developed for foreign languages, as far as the researcher‟s knowledge is concerned, there is no feature level opinion mining scheme for Amharic language. Therefore, this study proposes feature level opinion mining model for Amharic language by employing manually crafted rules and lexicon. The proposed model consists of five major components that can extract features, determine opinion words regarding identified features with their semantic orientation, aggregate multiple opinions and generate structured summary. Two experiments have been conducted for features extraction and opinion words determination by using 484 reviews from three different domains. The first experiment indicated that an average precision of 95.2% and recall of 26.1% were achieved in the features extraction and an average precision of 78.1% and recall of 66.8% were achieved in the determination of opinion words. The precision of the second experiment in features extraction gets lower by 15.4% whereas the precision of opinion words determination gets higher by 1.9% and the recall of both features extraction and opinion words determination gets higher by 7.8% and 25.9% respectively when compared to the first experiment. Thus, our experimental results demonstrate the effectiveness of the techniques we have applied.
Keywords: Opinion, Opinion Mining, Review, Blog, Semantic Orientation, Sentence Level, Document Level, Feature Level, Classification, Extraction, Determination.
Description
Keywords
Opinion, Opinion Mining, Review, Blog, Semantic Orientation, Sentence Level, Document Level, Feature Level, Classification, Extraction, Determination