Explainable Sentence-Level Sentiment Analysis for Amazon Product Reviews

TL;DR

This paper presents a sentence-level sentiment analysis approach for Amazon reviews using BiLSTM with attention, highlighting interpretability insights about aspect terms and sentiment words.

Contribution

It introduces an interpretable sentiment analysis model that analyzes attention weights at sentence and aspect levels, revealing interpretability insights.

Findings

Model achieves up to 96% accuracy.

Aspect terms receive equal or more attention than sentiment words.

Attention analysis enhances understanding of model interpretability.

Abstract

In this paper, we conduct a sentence level sentiment analysis on the product reviews from Amazon and thorough analysis on the model interpretability. For the sentiment analysis task, we use the BiLSTM model with attention mechanism. For the study of interpretability, we consider the attention weights distribution of single sentence and the attention weights of main aspect terms. The model has an accuracy of up to 0.96. And we find that the aspect terms have the same or even more attention weights than the sentimental words in sentences.