Elamur Research⌕
☰
home / papers / 2026 / NetGuardAI at MultiPRIDE: Multilingual Detection of Reclaimed Language

NetGuardAI at MultiPRIDE: Multilingual Detection of Reclaimed Language

Paper · 2026 · Noe Come Jacques Le Pollotec, Elena-Simona Apostol and Ciprian-Octavian Truică

Summary

The paper addresses a content moderation problem: distinguishing slurs used to attack LGBTQ+ people from slurs that community members reclaim for self-identification or solidarity. It describes the NetGuardAI team's entry to the MultiPRIDE shared task at EVALITA 2026, which covers tweets in Italian, English and Spanish. The task has two settings. In Task A, systems see only the tweet. In Task B, they may also use the author's profile biography.

The authors fine-tuned a compact multilingual transformer for binary classification. Non-reclaimed tweets outnumber reclaimed ones roughly 11 to 1, so they offset this with frequency-based class weighting in the loss. For Task B, tweet and bio were simply concatenated. An exploratory analysis using NMF topic modeling helps explain the language gaps:

  • Italian: reclaimed tweets are short and rich in first-person self-declarations.
  • English: reclaimed tweets often discuss the words themselves.

Italian Task A produced the best score, while English and Spanish lagged. Adding biographies reduced performance. The work offers an inexpensive baseline and shows where lightweight models break down on irony and meta-linguistic usage.

Key points

  • Model and training: distilbert-base-multilingual-cased with 6 layers and about 134M parameters. The [CLS] vector feeds a linear head with dropout 0.1. Inputs are capped at 128 tokens, and training ran for 2 epochs with AdamW on a Colab T4 GPU.
  • Task A results (Macro F1 on the test set): Italian 0.854, English 0.548, Spanish 0.648. The reclaimed-class F1 in English was only 0.187.
  • Task B results (Italian and Spanish only): Macro F1 of 0.740 and 0.493. The Spanish reclaimed-class F1 collapsed to 0.065, which the authors attribute to generic bio text acting as noise.
  • Exploratory analysis: Italian reclaimed tweets average 22.5 words, compared with 36.8 for hateful ones. Spanish reclaimed topics cluster around Pride events.
  • Code and future work: the code is released at github.com/DS4AI-UPB/NetGuardAI_MultiPride2026. Planned next steps include XLM-RoBERTa, minority-class data augmentation, and selective attention over biographies.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 2The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 works

🧭 Topics 1Research subjects and languages the work is about.Multilingual & cross-lingual NLP · 828 works

Cite

Cite as Noe Come Jacques Le Pollotec, Elena-Simona Apostol and Ciprian-Octavian Truică. 2026. NetGuardAI at MultiPRIDE: Multilingual Detection of Reclaimed Language. Proceedings of the Ninth Evaluation Campaign of Natural Language Processing and Speech Tools for Italian. Final Workshop (EVALITA 2026).

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{pollotec2026netguardai,
  title = {{NetGuardAI at MultiPRIDE: Multilingual Detection of Reclaimed Language}},
  author = {Noe Come Jacques {Le Pollotec}, Elena-Simona Apostol and Ciprian-Octavian Truică},
  booktitle = {Proceedings of the Ninth Evaluation Campaign of Natural Language Processing and Speech Tools for Italian. Final Workshop (EVALITA 2026)},
  year = {2026},
  url = {https://aclanthology.org/2026.evalita-1.32/}
}

References

29 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. E. U. A. for Fundamental Rights, Fundamental rights report 2024, 2024. URL: https://fra.europa.eu/ en/publication/2024/fundamental-rights-report-2024, last accessed: 2026-01-12. ↗
  2. C. Ferrando, L. Draetta, M. Madeddu, M. Sosto, V. Patti, P. Rosso, C. Bosco, J. Mata, E. Gualda, Multipride at evalita 2026: Overview of the multilingual automatic detection of slur reclamation in the lgbtq+ context task, in: Proceedings of the Ninth Evaluation Campaign of Natural Language Processin… find
  3. F. Cutugno, A. Miaschi, A. P. Aprosio, G. Rambelli, L. Siciliani, M. A. Stranisci, Evalita 2026: Overview of the 9th evaluation campaign of natural language processing and speech tools for italian, in: Proceedings of the Ninth Evaluation Campaign of Natural Language Processing and Speech Tools for I… find
  4. T. Davidson, D. Warmsley, M. Macy, I. Weber, Automated hate speech detection and the problem of offensive language, in: Proceedings of the 11th International AAAI Conference on Web and Social Media, 2017, pp. 512–515.
  5. J. Devlin, M.-W. Chang, K. Lee, K. Toutanova, Bert: Pre-training of deep bidirectional transformers for language understanding, in: Proceedings of NAACL-HLT, 2019.
  6. V.-I. Ilie, C.-O. Truică, E.-S. Apostol, A. Paschke, Context-Aware Misinformation Detection: A Benchmark of Deep Learning Architectures Using Word Embeddings, IEEE Access 9 (2021) 162122–162146. doi: 10.1109/ACCESS.2021.3132502 . — [doi:10.1109/access.2021.3132502](https://doi.org/10.1109/access.2021.3132502) find
  7. C.-O. Truică, E.-S. Apostol, It’s all in the Embedding! Fake News Detection using Document Embeddings, Mathematics 11 (2023) 1–29(508). doi: 10.3390/math11030508 . — [doi:10.3390/math11030508](https://doi.org/10.3390/math11030508) find
  8. C.-O. Truică, E.-S. Apostol, MisRoBÆRTa: Transformers versus Misinformation, Mathematics 10 (2022) 1–25(569). doi: 10.3390/math10040569 . — [doi:10.3390/math10040569](https://doi.org/10.3390/math10040569) find
  9. C.-O. Truică, E.-S. Apostol, A. Paschke, Awakened at CheckThat! 2022: Fake News Detection using BiLSTM and sentence transformer, in: Working Notes of the Conference and Labs of the Evaluation Forum, 2022, pp. 749–757. find
  10. M.-D. Cotelin, E.-S. Apostol, C.-O. Truică, NetGuardAI at EXIST2025: Sexism Detection using mDeBERTa, in: Working Notes of the Conference and Labs of the Evaluation Forum, 2025, pp. 1897–1905. find
  11. D.-E. Burghelea, C.-O. Truică, E.-S. Apostol, Verit-albert: A finetuned llm approach for verifying information credibility, in: The 2025 24th RoEduNet Conference: Networking in Education and Research (RoEduNet), 2025, pp. 1–6. doi: 10.1109/RoEduNet68395.2025.11208267 . — [doi:10.1109/roedunet68395.2025.11208267](https://doi.org/10.1109/roedunet68395.2025.11208267) find
  12. A. Petrescu, C.-O. Truică, E.-S. Apostol, Language-based Mixture of Transformers for EXIST2024, in: Working Notes of the Conference and Labs of the Evaluation Forum, 2024, pp. 1157–1164. find
  13. A. Petrescu, C.-O. Truică, E.-S. Apostol, Language-based Mixture of Transformers for Sexism Identification in Social Networks, in: Conference and Labs of the Evaluation Forum, 2025, pp. 142–155. doi: 10.1007/978-3-032-04354-2_10 . — [doi:10.1007/978-3-032-04354-2_10](https://doi.org/10.1007/978-3-032-04354-2_10) find
  14. A. Petrescu, E.-S. Apostol, C.-O. Truică, Awakened at EXIST2025: Adaptive Mixture of Transformers, in: Working Notes of the Conference and Labs of the Evaluation Forum, 2025, pp. 2112–2118. find
  15. Ciprian-Octavian, E.-S. Apostol, A.-G. Ilie, A. Paschke, HarmLLaMA: Harmful Language Detection with Large Language Models, in: International Conference on Intelligent Computer Communication and Processing (ICCP 2025), 2025. find
  16. C.-O. Truică, E.-S. Apostol, P. Karras, DANES: Deep Neural Network Ensemble Architecture for Social and Textual Context-aware Fake News Detection, Knowledge-Based Systems 294 (2024) 1–13(111715). doi: https://doi.org/10.1016/j.knosys.2024.111715 . — [doi:10.1016/j.knosys.2024.111715](https://doi.org/10.1016/j.knosys.2024.111715) ↗
  17. C.-O. Truică, E.-S. Apostol, M. Marogel, A. Paschke, GETAE: Graph Information Enhanced Deep Neural NeTwork Ensemble ArchitecturE for fake news detection, Expert Systems with Applications 275 (2025) 126984. doi: 10.1016/j.eswa.2025.126984 . — [doi:10.1016/j.eswa.2025.126984](https://doi.org/10.1016/j.eswa.2025.126984) find
  18. C.-O. Truică, E.-S. Apostol, T. S , tefu, P. Karras, A Deep Learning Architecture for Audience Interest Prediction of News Topic on Social Media, in: International Conference on Extending Database Technology (EDBT2021), 2021, pp. 588–599. doi: 10.5441/002/EDBT.2021.69 . — [doi:10.5441/002/edbt.2021.69](https://doi.org/10.5441/002/edbt.2021.69) find
  19. A. Petrescu, C.-O. Truică, E.-S. Apostol, P. Karras, Sparse shield: Social network immunization vs. harmful speech, in: ACM International Conference on Information & Knowledge Management, 2021, pp. 1426–1436. doi: 10.1145/3459637.3482481 . — [doi:10.1145/3459637.3482481](https://doi.org/10.1145/3459637.3482481) find
  20. C.-O. Truică, E.-S. Apostol, R.-C. Nicolescu, P. Karras, MCWDST: A Minimum-Cost Weighted Directed Spanning Tree Algorithm for Real-Time Fake News Mitigation in Social Media, IEEE Access 11 (2023) 125861–125873. doi: 10.1109/ACCESS.2023.3331220 . — [doi:10.1109/access.2023.3331220](https://doi.org/10.1109/access.2023.3331220) find
  21. E.-S. Apostol, Özgur Coban, C.-O. Truică, CONTAIN: A community-based algorithm for network immunization, Engineering Science and Technology, an International Journal 55 (2024) 1–10(101728). doi: https://doi.org/10.1016/j.jestch.2024.101728 . — [doi:10.1016/j.jestch.2024.101728](https://doi.org/10.1016/j.jestch.2024.101728) ↗
  22. A. Petrescu, C.-O. Truică, E.-S. Apostol, A. Paschke, EDSA-Ensemble: An Event Detection Sentiment Analysis Ensemble Architecture, IEEE Transactions on Affective Computing 16 (2025) 555–572. doi: 10.1109/taffc.2024.3434355 . — [doi:10.1109/taffc.2024.3434355](https://doi.org/10.1109/taffc.2024.3434355) find
  23. E.-S. Apostol, C.-O. Truică, A. Paschke, ContCommRTD: A Distributed Content-Based Misinformation-Aware Community Detection System for Real-Time Disaster Reporting, IEEE Transactions on Knowledge and Data Engineering (2024) 1–12. doi: 10.1109/tkde.2024. 3417232 . — [doi:10.1109/tkde.2024](https://doi.org/10.1109/tkde.2024) find
  24. C.-O. Truică, A.-T. Constantinescu, E.-S. Apostol, StopHC: A Harmful Content Detection and Mitigation Architecture for Social Media Platforms, in: IEEE International Conference on Intelligent Computer Communication and Processing (ICCP 2024), 2024, pp. 1–5. doi: 10.1109/ICCP63557. 2024.10793051 . — [doi:10.1109/iccp63557](https://doi.org/10.1109/iccp63557) find
  25. M.-D. Cotelin, C.-O. Truică, E.-S. Apostol, Cleannews: a network-aware fake news mitigation architecture for social media, arXiv preprint arXiv:2509.04489 (2025). — [arXiv:2509.04489](https://arxiv.org/abs/2509.04489) find
  26. C.-O. Truică, J. Darmont, J. Velcin, A scalable document-based architecture for text analysis, in: International Conference on Advanced Data Mining and Applications, Springer, 2016, pp. 481–494. doi: 10.1007/978-3-319-49586-6_33 . — [doi:10.1007/978-3-319-49586-6_33](https://doi.org/10.1007/978-3-319-49586-6_33) find
  27. C.-O. Truica, A. Guille, M. Gauthier, Cats: Collection and analysis of tweets made simple, in: ACM Conference on Computer Supported Cooperative Work and Social Computing Companion, ACM, 2016, pp. 41–44. doi: 10.1145/2818052.2874320 . — [doi:10.1145/2818052.2874320](https://doi.org/10.1145/2818052.2874320) find
  28. V. Sanh, L. Debut, J. Chaumond, T. Wolf, Distilbert, a distilled version of bert: smaller, faster, cheaper and lighter, arXiv preprint arXiv:1910.01108 (2019). find
  29. C.-O. Truică, E. S. Apostol, C. A. Leordeanu, Topic modeling using contextual cues, in: IEEE International Symposium on Symbolic and Numeric Algorithms for Scientific Computing (SYNASC2017), 2017, pp. 203–210. doi: 10.1109/synasc.2017.00041 . — [doi:10.1109/synasc.2017.00041](https://doi.org/10.1109/synasc.2017.00041) find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/pollotec-2026-netguardai-multipride-multilingual-detection-reclaimed/