الفهرس | Only 14 pages are availabe for public view |
Abstract Through the state-of-the-art digitalization, we can see a massive growth in user-generated content on the web that provides feedback from people on a variety of topics. However, manually managing largescale user feedback would be a hard task and a waste of time. Therefore, the concept of opinion mining or sentiment analysis is emerged. Opinion mining is a computerized study of individuals’ feelings and opinions about an entity or product. Opinion mining is difficult because the users’ opinions or reviews are mostly unstructured text, lower quality, noisy, and spam. Thus, there are several challenges facing the opinion mining and causing poor classification performance. Some of these challenges are multiple languages of user reviews, fake reviews, manipulation of emoticons, implicit aspects, spam aspects, and negative words that change the class label of opinion words. In this thesis a semantic-based aspect level opinion mining (SALOM) model is proposed. In order to address some of the previously mentioned challenges. Whereas, SALOM involves the semantic similarity measure and Wordnet ontology to extract the product related domain aspects. Moreover, it considers other types of product aspects such as aspect-related synonym, hyponym, and hypernym. Not only that, but the proposed model also considers the reviews that contain product aspects only. In addition, SALOM model considers negation words that affect the performance of the opinion mining process. Five different datasets are used to evaluate the proposed SALOM model. SALOM achieved promising results compared to other methods. Where, the performance reached 91.6% for accuracy, 93% for recall, 95.8% for precision, and 93.3% for f-measure. |