Deep Learning Brasil at ABSAPT 2022: Portuguese Transformer Ensemble Approaches

TL;DR

This paper presents the best performing systems for Portuguese Aspect-Based Sentiment Analysis at ABSAPT 2022, achieving new state-of-the-art results on aspect term extraction and sentiment polarity classification.

Contribution

It introduces ensemble approaches that significantly improve performance in Portuguese ABSA tasks, setting new benchmarks.

Findings

Achieved state-of-the-art results on ATE and SOE subtasks

Ensemble methods outperform previous models

Contributed to Portuguese sentiment analysis research

Abstract

Aspect-based Sentiment Analysis (ABSA) is a task whose objective is to classify the individual sentiment polarity of all entities, called aspects, in a sentence. The task is composed of two subtasks: Aspect Term Extraction (ATE), identify all aspect terms in a sentence; and Sentiment Orientation Extraction (SOE), given a sentence and its aspect terms, the task is to determine the sentiment polarity of each aspect term (positive, negative or neutral). This article presents we present our participation in Aspect-Based Sentiment Analysis in Portuguese (ABSAPT) 2022 at IberLEF 2022. We submitted the best performing systems, achieving new state-of-the-art results on both subtasks.