Papers by Farah Benamara
CrisisTS: Coupling Social Media Textual Data and Meteorological Time Series for Urgency Classification (2025.acl-long)
Copied to clipboard
| Challenge: | Existing studies on fusion of texts and tabular-based time series to improve performance of NLP applications have focused on coupling texts with tabular time series. |
| Approach: | They propose a multimodal and multilingual dataset for urgency classification that allows for temporal and location alignment even in the absence of location mention in the text. |
| Outcome: | The proposed dataset outperforms text-only models in many applications while ensuring model generalizability. |
Image and Text: Fighting the same Battle? Super Resolution Learning for Imbalanced Text Classification (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Using high-resolution images to overcome the problem of low resolution has never been used in NLP. |
| Approach: | They propose a super-resolution learning method that uses high-res images to overcome the problem of low resolution images. |
| Outcome: | The proposed method is efficient when compared to state-of-the-art methods on several benchmarks datasets in two languages. |
Are Dialects Better Prompters? A Case Study on Arabic Subjective Text Classification (2025.findings-acl)
Copied to clipboard
| Challenge: | specialized fine-tuned models with Arabic and Arabizi scripts achieve the best results . transfer learning demonstrated limited effectiveness, despite ability to adapt to other regionally specific varieties . |
| Approach: | They evaluate the performance of 12 open source LLMs for Arabic and Arabizi scripts . they highlight the impact of Arabic-centric LLM fine-tuning and prompt design on models . |
| Outcome: | The results show that specialized models with Arabic and Arabizi scripts achieve the best results . |
Speech acts and Communicative Intentions for Urgency Detection (2022.starsem-1)
Copied to clipboard
| Challenge: | Existing approaches to detect speech acts (SA) in synchronous and asynchronous dialogues have been proposed to capture communicative intentions on the part of the speaker. |
| Approach: | They propose to annotate tweets with urgency and SA and develop deep learning architectures to inject it into urgency detection. |
| Outcome: | The proposed dataset annotated for urgency and SA improves information type detection in an out-of-type configuration where models are evaluated in unseen event types during training. |
Humans Need Context, What about Machines? Investigating Conversational Context in Abusive Language Detection (2024.lrec-main)
Copied to clipboard
| Challenge: | In this paper, we examine the role of conversational context in abusive language detection . prior studies have ignored the contextual nature of abusive language, ignoring this aspect . toxicity, hate speech, harmful stereotypes are among the forms of harmful language . |
| Approach: | They propose to use conversational context to analyze abusive language detection using two methods . they use "abusive language" as an umbrella term to refer to various forms of harmful language . |
| Outcome: | The proposed approach is based on two datasets in English and a new dataset of French tweets annotated for hate speech and stereotypes. |
What Did You Learn To Hate? A Topic-Oriented Analysis of Generalization in Hate Speech Detection (2023.eacl-main)
Copied to clipboard
| Challenge: | Hate speech detection datasets often use different annotation guidelines, resulting in inconsistencies . authors propose a topic-oriented approach to study generalization across popular hate speech datasets . |
| Approach: | They propose a topic-oriented approach to study generalization across popular hate speech datasets . they compare Transformer-based models in capturing topic-generic and topic-specific knowledge . |
| Outcome: | The proposed approach improves the reliability of hate speech detection on social media platforms. |
Can LLM Safety Be Ensured by Constraining Parameter Regions? (2026.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) are often assumed to contain parameter subsets whose modification directly influences safety behaviors. |
| Approach: | They evaluate four methods to identify parameter subsets with "safety regions" they find low overlap, but overlap drops when refinement is done using utility datasets . |
| Outcome: | The proposed methods show low overlap and drop significantly when refined using utility datasets. |
EIFFEL: a novel benchmark to measure bias of English heavy training on French idiomatic expressions (2026.acl-long)
Copied to clipboard
| Challenge: | Mainstream multilingual models are generally trained on a much higher proportion of English data . this raises questions about their ability to capture linguistic features specific to non-English languages . |
| Approach: | They propose a benchmark to test multilingual LLMs' ability to capture linguistic features in other languages. |
| Outcome: | The proposed benchmark shows that multilingual models can capture features in non-English languages and cultural norms. |
An Annotated Corpus for Sexism Detection in French Tweets (2020.lrec-1)
Copied to clipboard
Patricia Chiril, Véronique Moriceau, Farah Benamara, Alda Mari, Gloria Origgi, Marlène Coulomb-Gully
| Challenge: | Social media networks allow users to share opinions and sentiments, which can cause a large spreading of hatred or abusive messages. |
| Approach: | They propose to annotate 12,000 tweets with a sexism detection scheme in France . they propose to use deep learning to detect if a message with sexist content is really s. |
| Outcome: | The proposed scheme detects sexist content and identifies if it is really sexism . the proposed scheme is the first of its kind in the u.s. |
Give me your Intentions, I’ll Predict our Actions: A Two-level Classification of Speech Acts for Crisis Management in Social Media (2022.lrec-1)
Copied to clipboard
| Challenge: | Using social networks, social media is a vital tool for emergency management and social media has been used to generate valuable information in crisis situations. |
| Approach: | They propose to measure for the first time the role of SA on urgency detection in tweets . they propose to use a two-layer annotation scheme to annotate tweets for both SA and urgency . |
| Outcome: | The proposed scheme combines two-layer annotation scheme and deep learning experiments to detect SA in a crisis corpus. |
Automatic Detection of Stigmatizing Uses of Psychiatric Terms on Twitter (2022.lrec-1)
Copied to clipboard
| Challenge: | Psychiatry and people suffering from mental disorders have often been given a pejorative label that induces social rejection. |
| Approach: | They propose to use deep learning to detect polarity and type of use in tweets . they propose to combine polarization detection with typeof use detection to improve polarities . |
| Outcome: | The proposed models can detect the polarity of a tweet and the types of use on a dataset that is not yet available. |
A Multilingual Dataset of Racial Stereotypes in Social Media Conversational Threads (2023.findings-eacl)
Copied to clipboard
Tom Bourgeade, Alessandra Teresa Cignarella, Simona Frenda, Mario Laurent, Wolfgang Schmeisser-Nieto, Farah Benamara, Cristina Bosco, Véronique Moriceau, Viviana Patti, Mariona Taulé
| Challenge: | a new corpus-based study addresses racial stereotypes in social media conversations . a multilingual corpus of rhs is used to investigate how they are spread . |
| Approach: | They propose a corpus-based method for multilingual racial stereotype identification in social media conversational threads. |
| Outcome: | The proposed method sheds light on how racial hoaxes are spread and allows identification of negative stereotypes that reinforce them. |
Multilingual Irony Detection with Dependency Syntax and Neural Models (2020.coling-main)
Copied to clipboard
Alessandra Teresa Cignarella, Valerio Basile, Manuela Sanguinetti, Cristina Bosco, Paolo Rosso, Farah Benamara
| Challenge: | Several semantic and syntactic devices can be used to express irony, causing the incongruity, determine the clash and play the role of irony triggers within a text. |
| Approach: | They propose to exploit linguistic resources where syntax is annotated according to the Universal Dependencies scheme. |
| Outcome: | The proposed method exploits linguistic resources where syntax is annotated according to the Universal Dependencies scheme. |
An Algerian Corpus and an Annotation Platform for Opinion and Emotion Analysis (2020.lrec-1)
Copied to clipboard
| Challenge: | Currently, there are more than 4 billion Internet users worldwide . the number of social media users in Algeria has tripled over a year . |
| Approach: | They propose a platform for crowdsourcing annotation of tweets at different levels of granularity. |
| Outcome: | The proposed platform can be used to create the largest Algerian dialect subjectivity lexicon of about 9,000 entries. |
“Be nice to your wife! The restaurants are closed”: Can Gender Stereotype Detection Improve Sexism Classification? (2021.findings-emnlp)
Copied to clipboard
| Challenge: | a new study examines the impact of gender stereotype detection on sexism classification . GS is defined as "pictures in our heads" and is used to describe social group members . |
| Approach: | They propose to use tweets as a dataset to detect sexist hate speech . they propose a method for data augmentation based on sentence similarity with external datasets . |
| Outcome: | The proposed method detects sexist hate speech in tweets and then uses it for sexism classification. |
CDB: A Unified Framework for Hope Speech Detection Through Counterfactual, Desire and Belief (2025.findings-naacl)
Copied to clipboard
Tulio Ferreira Leite Da Silva, Gonzalo Freijedo Aduna, Farah Benamara, Alda Mari, Zongmin Li, Li Yue, Jian Su
| Challenge: | Using algorithms to model user-generated desires on social media, we propose a new approach to understanding and detection of hope speech. |
| Approach: | They propose a language-driven decomposition of the notional category hope and its automatic detection in a unified setting. |
| Outcome: | The proposed model captures future-oriented hopes through desires and beliefs and the counterfactuality of past unfulfilled wishes and regrets. |
He said “who’s gonna take care of your children when you are at ACL?”: Reported Sexist Acts are Not Sexist (2020.acl-main)
Copied to clipboard
Patricia Chiril, Véronique Moriceau, Farah Benamara, Alda Mari, Gloria Origgi, Marlène Coulomb-Gully
| Challenge: | Sexism is prejudice or discrimination based on a person's gender. |
| Approach: | They propose to use a French dataset annotated for sexism detection to characterize sexist content and to train deep learning experiments on tweets. |
| Outcome: | The proposed dataset is the first to be used for sexism detection in France and constitutes a first step towards offensive content moderation. |
How’s Business Going Worldwide ? A Multilingual Annotated Corpus for Business Relation Extraction (2022.lrec-1)
Copied to clipboard
| Challenge: | The 21st century economy has shaped the economic landscape and changed the way market stakeholders interact with each other in the global market where national borders have melted and trades became more open and free. |
| Approach: | They propose a multilingual dataset for automatic extraction of binary business relations involving organizations from the web. |
| Outcome: | The proposed dataset is the first multilingual dataset for automatic extraction of binary business relations involving organizations from the web. |