Multilingual and Multilabel Emotion Recognition using Virtual Adversarial Training
This work addresses emotion recognition in multiple languages and labels, which is incremental as it extends an existing method to a new task.
The paper tackles multilingual and multilabel emotion recognition by applying Virtual Adversarial Training (VAT) to leverage unlabeled data across languages, resulting in performance gains of up to 6.2% over supervised learning and improvements of up to 7% in state-of-the-art Jaccard Index scores for languages like Arabic and Spanish.
Virtual Adversarial Training (VAT) has been effective in learning robust models under supervised and semi-supervised settings for both computer vision and NLP tasks. However, the efficacy of VAT for multilingual and multilabel text classification has not been explored before. In this work, we explore VAT for multilabel emotion recognition with a focus on leveraging unlabelled data from different languages to improve the model performance. We perform extensive semi-supervised experiments on SemEval2018 multilabel and multilingual emotion recognition dataset and show performance gains of 6.2% (Arabic), 3.8% (Spanish) and 1.8% (English) over supervised learning with same amount of labelled data (10% of training data). We also improve the existing state-of-the-art by 7%, 4.5% and 1% (Jaccard Index) for Spanish, Arabic and English respectively and perform probing experiments for understanding the impact of different layers of the contextual models.