PROJECT TITLE :

Aspect-Context Interactive Attention Representation for Aspect-Level Sentiment Classification

ABSTRACT:

The goal of aspect-level sentiment classification is to determine the sentiment polarities of distinct aspects in reviews, where each review often comprises many aspects with varying polarities. Distinct context representations are required for different aspects in aspect-level sentiment classification, as opposed to document-level sentiment classification. Existing techniques typically model aspects and contexts independently using the Long Short-Term Memory (LSTM) network, then integrate Attention Mechanisms to extract properties of a given aspect in its context. Because attention mechanisms aren't used in sequence modeling, context sequence representations aren't taken into account. This paper offers a novel aspect-context interactive representation structure that generates sequence-to-sequence representations in both context and aspect using simply an Attention Mechanism. During the context sequence modeling process, it can extract features relevant to a given aspect and generate a high-quality aspect representation at the same time. We did extensive tests to compare our method to thirteen other ways. The suggested model achieves much better results on the Restaurant dataset, as well as extremely competitive results on the Laptop and Twitter datasets, according to our experiments.