Assessing Linguistic Generalisation in Language Models: A Dataset for Brazilian Portuguese

TL;DR

This paper introduces a new dataset and intrinsic evaluation tasks to assess how well language models for Brazilian Portuguese capture linguistic phenomena like grammatical structures and multiword expressions, aiding transparency and comparability.

Contribution

It presents a novel dataset and evaluation framework specifically designed to analyze linguistic generalisation in Brazilian Portuguese language models.

Findings

BERTimbau Large outperforms BERTimbau Base and mBERT on MWE tasks.

The dataset effectively distinguishes models' capabilities in capturing linguistic phenomena.

Evaluation reveals strengths and limitations of current models in encoding grammatical structures.

Abstract

Much recent effort has been devoted to creating large-scale language models. Nowadays, the most prominent approaches are based on deep neural networks, such as BERT. However, they lack transparency and interpretability, and are often seen as black boxes. This affects not only their applicability in downstream tasks but also the comparability of different architectures or even of the same model trained using different corpora or hyperparameters. In this paper, we propose a set of intrinsic evaluation tasks that inspect the linguistic information encoded in models developed for Brazilian Portuguese. These tasks are designed to evaluate how different language models generalise information related to grammatical structures and multiword expressions (MWEs), thus allowing for an assessment of whether the model has learned different linguistic phenomena. The dataset that was developed for these tasks is composed of a series of sentences with a single masked word and a cue phrase that helps in narrowing down the context. This dataset is divided into MWEs and grammatical structures, and the latter is subdivided into 6 tasks: impersonal verbs, subject agreement, verb agreement, nominal agreement, passive and connectors. The subset for MWEs was used to test BERTimbau Large, BERTimbau Base and mBERT. For the grammatical structures, we used only BERTimbau Large, because it yielded the best results in the MWE task.