EXPLORATION OF AN ONLINE PRODUCT EVALUATION SYSTEM UTILIZING RATINGS AND REVIEW COMMENTS

COMPUTER SCIENCE

Mining opinion targets and opinion words from online reviews are critical tasks for fine-grained opinion mining. The fundamental component of this process is finding relationships between words that express opinions. In this work, we propose a unique technique based on the partially supervised alignment model, which treats the detection of opinion relations as an alignment process. We then utilize a graph-based co-ranking technique to estimate the confidence of each candidate. Finally, we identify the candidates with the highest confidence as opinion targets or opinion words. Our approach allows for more precise capture of opinion relations compared to previous techniques based on nearest neighbor principles, particularly for long-span connections. When dealing with informal online documents, our word alignment approach effectively mitigates the negative impact of parsing mistakes, unlike syntax-based solutions. By incorporating partial supervision, our suggested model outperforms typical unsupervised alignment models in terms of precision. Furthermore, in our graph-based co-ranking approach, while assessing candidate confidence, we penalize higher-degree vertices to reduce the likelihood of creating mistakes. Through experiments conducted on three corpora of varying sizes and languages, we have demonstrated that our method outperforms state-of-the-art approaches

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.