SINAI at SemEval-2021 Task 5: Combining Embeddings in a BiLSTM-CRF model for Toxic Spans Detection

Proceedings of the 15th International Workshop on Semantic Evaluation (SemEval-2021), pages 984–989 Bangkok, Thailand (online), August 5–6, 2021. ©2021 Association for Computational Linguistics 984 SINAI at SemEval-2021 Task 5: Combining Embeddings in a BiLSTM-CRF model for Toxic Spans Detection Flor Miriam Plaza-del-Arco, Pilar L´opez- ´Ubeda L. Alfonso Ure˜na-L´opez, M. Teresa Mart´ın-Valdivia Department of Computer Science, Advanced Studies Center in ICT (CEATIC) Universidad de Ja´en, Campus Las Lagunillas, 23071, Ja´en, Spain {fmplaza, plubeda, laurena, maite}@ujaen.es Abstract This paper describes the participation of SINAI team at Task 5: Toxic Spans Detec- tion which consists of identifying spans that make a text toxic. Although several resources and systems have been developed so far in the context of offensive language, both annotation and tasks have mainly focused on classifying whether a text is offensive or not. However, detecting toxic spans is crucial to identify why a text is toxic and can assist human modera- tors to locate this type of content on social me- dia. In order to accomplish the task, we follow a deep learning-based approach using a Bidi- rectional variant of a Long Short Term Mem- ory network along with a stacked Conditional Random Field decoding layer (BiLSTM-CRF). Specifically, we test the performance of the combination of different pre-trained word em- beddings for recognizing toxic entities in text. The results show that the combination of word embeddings helps in detecting offensive con- tent. Our team ranks 29th out of 91 partici- pants. 1 Introduction The advance of online communication has in- creased the use of offensive or toxic language in several websites, including social networks such as Instagram, Twitter, or YouTube. Consequently, this type of prejudiced communication could lead to negative psychological effects among Internet users, causing anxiety, harassment, and even sui- cide in extreme cases (Hinduja and Patchin, 2010). Moderation is essential to promote healthy on- line communication. Therefore, governments, on- line communities, and social media platforms are continuously taking appropriate actions to imple- ment laws and policies combating toxic language on the Web. In order to help to track this type of comments and due to the amount of data generated every day on the Web, automatic systems based on Natural Language Processing (NLP) techniques are required. In particular, offensive language de- tection and analysis has become an important area of research in NLP, resulting in several studies that are contributing to combating this website phe- nomenon (Plaza-del Arco et al., 2019; Zampieri et al., 2019a; Ranasinghe et al., 2019; Plaza del Arco et al., 2020). In this paper1, we present our proposal system as part of our participation in SemEval-2021 Task 5: Toxic Spans Detection (Pavlopoulos et al., 2021), which aims to identify entities that refer to a toxic language in the text. To accomplish the task, our team focused on detecting specific types of toxic en- tities in the text using a methodology based on the BiLSTM-CRF model showing that the combina- tion of different pre-trained language embeddings succeeds in detecting toxic entities. The rest of the paper …

Content truncated to 500 words