Divide (Text) and Conquer (Sentiment): Improved Sentiment Classification by Constituent Conflict Resolution

TL;DR

This paper presents a new approach to sentiment classification that isolates conflicting sentiments within passages and aggregates them using an MLP, significantly improving accuracy and efficiency over traditional methods.

Contribution

It introduces novel methodologies for resolving constituent conflicts in passages and demonstrates the effectiveness of an MLP-based aggregation strategy.

Findings

MLP aggregation outperforms baseline models

Method reduces computational cost by approximately 99%

Effective across multiple datasets including Amazon, Twitter, and SST

Abstract

Sentiment classification, a complex task in natural language processing, becomes even more challenging when analyzing passages with multiple conflicting tones. Typically, longer passages exacerbate this issue, leading to decreased model performance. The aim of this paper is to introduce novel methodologies for isolating conflicting sentiments and aggregating them to effectively predict the overall sentiment of such passages. One of the aggregation strategies involves a Multi-Layer Perceptron (MLP) model which outperforms baseline models across various datasets, including Amazon, Twitter, and SST while costing $\sim$1/100 of what fine-tuning the baseline would take.