Abstract :

Among the several tasks that make up aspect-based sentiment analysis (ABSA), aspect identification stands out as the most important.It is a challenging problem to address because of the subjectivity of classification and the fact that classes often overlap. The bulk of the machine learning algorithms utilized to address ACD have been statistical behavior-based rule-based strategies. The approach used in this article is based on association rules. To overcome association rules' statistical shortcomings, we devised a mixed-principles approach that combines association rule mining with semantic associations. We used word-embed as a means of establishing semantic connections. The SemEval dataset, a benchmark for feature classification in the industry, was used to conduct the studies. We found that, in addition to statistical relationships, semantic links might improve classification accuracy. In comparison to other statistical approaches, the one put out fared better.