To BAN or not to BAN : Bayesian attention networks for reliable hate speech detection

Miok, Kristian (avtor), Škrlj, Blaž (avtor), Zaharie, Daniela (avtor), Robnik Šikonja, Marko (avtor)

Hate speech is an important problem in the management of user-generated content. To remove offensive content or ban misbehaving users, content moderators need reliable hate speech detectors. Recently, deep neural networks based on the transformer architecture, such as the (multilingual) BERT model, have achieved superior performance in many natural language classification tasks, including hate speech detection. So far, these methods have not been able to quantify their output in terms of reliability. We propose a Bayesian method using Monte Carlo dropout within the attention layers of the transformer models to provide well-calibrated reliability estimates. We evaluate and visualize the results of the proposed approach on hate speech detection problems in several languages. Additionally, we test whether affective dimensions can enhance the information extracted by the BERT model in hate speech classification. Our experiments show that Monte Carlo dropout provides a viable mechanism for reliability estimation in transformer networks. Used within the BERT model, it offers state-of-the-art classification performance and can detect less trusted predictions.

• Jezik: Angleški jezik
• Ključne besede: natural language processing, machine learning, transformer neural networks, Bayesian neural networks, BERT models, prediction uncertainty, reliability estimation, Monte Carlo dropout, Bayesian BERT, sentic computing, model calibration
• Vrsta gradiva: Članek v reviji
• Tipologija: 1.01 - Izvirni znanstveni članek
• Organizacija: FRI - Fakulteta za računalništvo in informatiko
• Status publikacije: Objavljeno
• Različica publikacije: Objavljena publikacija
• Leto izida: 2022
• Št. strani: Str. 353-371
• Številčenje: Vol. 14, iss. 1
• PID: 20.500.12556/RUL-144813
• UDK: 004.85:81'322.2
• ISSN pri članku: 1866-9956
• DOI: 10.1007/s12559-021-09826-9
• COBISS.SI-ID: 80879363
• Datum objave v RUL: 14.03.2023

Gradivo je del revije

• Naslov: Cognitive computation
• Skrajšan naslov: Cogn. comput.
• Založnik: Springer Nature
• ISSN: 1866-9956
• COBISS.SI-ID: 80861443

Licence

• Licenca: CC BY 4.0, Creative Commons Priznanje avtorstva 4.0 Mednarodna
• Povezava: http://creativecommons.org/licenses/by/4.0/deed.sl
• Opis: To je standardna licenca Creative Commons, ki daje uporabnikom največ možnosti za nadaljnjo uporabo dela, pri čemer morajo navesti avtorja.

Sekundarni jezik

• Jezik: Slovenski jezik
• Ključne besede: obdelava naravnega jezika, strojno učenje, nevronske mreže transformer, bayesovske nevronske mreže, modeli BERT

Projekti

• Financer: ARRS - Agencija za raziskovalno dejavnost Republike Slovenije
• Številka projekta: P6-0411
• Naslov: Jezikovni viri in tehnologije za slovenski jezik

• Financer: EC - European Commission
• Program financ.: H2020
• Številka projekta: 825153
• Naslov: Cross-Lingual Embeddings for Less-Represented Languages in European News Media
• Akronim: EMBEDDIA