The MOL consists of the first specialized lexicon for hate speech detection, which was annotated with contextual information. It is composed of 1,000 explicit and implicit terms and expressions with pejorative connotations, which were manually identified by a linguist and annotated by three different annotators with contextual information (context-dependent or context-independent). For example, the term "stupid" consists of a context-independent offensive term since this term is mostly found in pejorative context of use. Differently, the terms "useless" and "worm" are classified as context-dependent offensive terms because they may be found in both contexts of use: (i) non-pejorative connotation, such as "this smartphone is useless" or "the fisherman uses worms for bait", as well as (ii) pejorative connotation, such as "this last President was useless" or "this being human is such a worm".
Multilingual Offensive Lexicon was extracted manually by a linguist from the HateBR dataset, and each term and expression were annotated by three different annotators, obtaining a high human-agreement score (73% Kappa). MOL was originally written in Portuguese and manually translated by native speakers takes into consideration their cultural adaptations in English, Spanish, French, German, and Turkish. Therefore, MOL is available in six different languages.
The table below describes the MOL statistics
| Contextual Information | Hate Targets | ||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
|
Vargas, F., Carvalho, I., Pardo, T.A.S., Benevenuto, F. (2024). Context-Aware and Expert Data Resources for Brazilian Portuguese Hate Speech Detection. Natural Language Processing Journal - Cambridge Core. pp.1-21.
Vargas, F., Góes, F., Carvalho, I., Benevenuto, F., Pardo, T.A.S. (2022). Contextual-Lexicon Approach for Abusive Language Detection. Proceedings of the 13th International Conference on Recent Advances in Natural Language Processing (RANLP 2021). pp.1438–1447. Held Online. https://aclanthology.org/2021.ranlp-1.161
@inproceedings{vargas-etal-2021-contextual, title = "Contextual-Lexicon Approach for Abusive Language Detection", author = "Vargas, Francielle and Rodrigues de G{\'o}es, Fabiana and Carvalho, Isabelle and Benevenuto, Fabr{\'\i}cio and Pardo, Thiago", editor = "Mitkov, Ruslan and Angelova, Galia", booktitle = "Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2021)", year = "2021", address = "Held Online", publisher = "INCOMA Ltd.", url = "https://aclanthology.org/2021.ranlp-1.161", pages = "1438--1447", }
