[PDF][PDF] Haspeede 2@ evalita2020: Overview of the evalita 2020 hate speech detection task

M Sanguinetti, G Comandini, E Di Nuovo… - … Campaign of Natural …, 2020 - research.rug.nl
Abstract The Hate Speech Detection (HaSpeeDe 2) task is the second edition of a shared
task on the detection of hateful content in Italian Twitter messages. HaSpeeDe 2 is …

[PDF][PDF] AMI@ EVALITA2020: Automatic misogyny identification

E Fersini, D Nozza, P Rosso - … of the 7th evaluation campaign of …, 2020 - iris.unibocconi.it
Automatic Misogyny Identification (AMI) is a shared task proposed at the Evalita 2020
evaluation campaign. The AMI challenge, based on Italian tweets, is organized into two …

Entropy-based attention regularization frees unintended bias mitigation from lists

G Attanasio, D Nozza, D Hovy, E Baralis - arxiv preprint arxiv:2203.09192, 2022 - arxiv.org
Natural Language Processing (NLP) models risk overfitting to specific terms in the training
data, thereby reducing their performance, fairness, and generalizability. Eg, neural hate …

Lost in translation: large language models in non-English content analysis

G Nicholas, A Bhatia - arxiv preprint arxiv:2306.07377, 2023 - arxiv.org
In recent years, large language models (eg, Open AI's GPT-4, Meta's LLaMa, Google's
PaLM) have become the dominant approach for building AI systems to analyze and …

[PDF][PDF] Benchmarking post-hoc interpretability approaches for transformer-based misogyny detection

G Attanasio, D Nozza, E Pastor… - Proceedings of NLP …, 2022 - iris.unibocconi.it
Warning: This paper contains examples of language that some people may find offensive.
Transformer-based Natural Language Processing models have become the standard for …

Capturing covertly toxic speech via crowdsourcing

A Lees, D Borkan, I Kivlichan, J Nario… - Proceedings of the First …, 2021 - aclanthology.org
We study the task of labeling covert or veiled toxicity in online conversations. Prior research
has highlighted the difficulty in creating language models that recognize nuanced toxicity …

[PDF][PDF] MilaNLP at semeval-2022 task 5: Using perceiver IO for detecting misogynous memes with text and image modalities

G Attanasio, D Nozza, F Bianchi - Proceedings of the 16th …, 2022 - iris.unibocconi.it
Warning: This paper contains examples of language that some people may find offensive or
upsetting. In this paper, we describe the system proposed by the MilaNLP team for the …

HATE-ITA: Hate speech detection in Italian social media text

D Nozza, F Bianchi, G Attanasio - … of the Sixth Workshop on Online …, 2022 - aclanthology.org
Online hate speech is a dangerous phenomenon that can (and should) be promptly
counteracted properly. While Natural Language Processing supplies appropriate algorithms …

Misogyny and aggressiveness tend to come together and together we address them

A Muti, F Fernicola… - Proceedings of the …, 2022 - aclanthology.org
We target the complementary binary tasks of identifying whether a tweet is misogynous and,
if that is the case, whether it is also aggressive. We compare two ways to address these …

A checkpoint on multilingual misogyny identification

A Muti, A Barrón-Cedeño - … of the 60th Annual Meeting of the …, 2022 - aclanthology.org
We address the problem of identifying misogyny in tweets in mono and multilingual settings
in three languages: English, Italian, and Spanish. We explore model variations considering …