ABSAPT 2022 at IberLEF: Overview of the Task on Aspect-Based Sentiment Analysis in Portuguese

Felix L. V. da Silva, Guilherme da S. Xavier, Heliks M. Mensenburg, Rodrigo F. Rodrigues, Leonardo P. dos Santos, Ricardo M. Araújo, Ulisses B. Corrêa, Larissa A. de Freitas

Resumen


This paper presents the task on Aspect-Based Sentiment Analysis in Portuguese (ABSAPT), held within Iberian Languages Evaluation Forum (IberLEF 2022). We asked the participants to develop systems capable of extracting aspects (AE) and classifying sentiment of aspects (ASC) in texts. We created one corpora containing reviews about hotels. Twelve teams registered to the task, among which five submitted predictions and technical reports. The best performing system achieved an Accuracy (Acc) value of 0.67 in AE sub-task (Team Deep Learning Brasil) and a Balanced Accuracy (Bacc) value of 0.82 in ASC sub-task (Team Deep Learning Brasil).

Texto completo:

PDF