no code implementations • NAACL 2022 • Eyup Yilmaz, Cagri Toraman
Supervised training with cross-entropy loss implicitly forces models to produce probability distributions that follow a discrete delta distribution.
1 code implementation • 13 May 2024 • Cagri Toraman
Despite advancements in English-dominant generative large language models, further development is needed for low-resource languages to enhance global accessibility.
no code implementations • 3 Apr 2024 • Arianna Muti, Federico Ruggeri, Cagri Toraman, Lorenzo Musetti, Samuel Algherini, Silvia Ronchi, Gianmarco Saretto, Caterina Zapparoli, Alberto Barrón-Cedeño
Disambiguating the meaning of such terms might help the detection of misogyny.
no code implementations • 27 Jul 2023 • Izzet Emre Kucukkaya, Umitcan Sahin, Cagri Toraman
For the easy and medium setups, we submit transition-focused natural language inference based on DeBERTa with warmup training, and the same model without transition for the hard setup.
no code implementations • 27 Jul 2023 • Umitcan Sahin, Izzet Emre Kucukkaya, Cagri Toraman
In this paper, we describe our approach for the Trigger Detection shared task at PAN CLEF 2023, where we want to detect multiple triggering content in a given Fanfiction document.
no code implementations • 25 Jul 2023 • Umitcan Sahin, Izzet Emre Kucukkaya, Oguzhan Ozcelik, Cagri Toraman
Throughout the Russia-Ukraine war, both opposing factions heavily relied on text-embedded images as a vehicle for spreading propaganda and hate speech.
1 code implementation • 26 Feb 2023 • Cagri Toraman, Izzet Emre Kucukkaya, Oguzhan Ozcelik, Umitcan Sahin
The importance of social media is again exposed in the recent tragedy of the 2023 Turkey and Syria earthquake.
1 code implementation • 11 Oct 2022 • Cagri Toraman, Oguzhan Ozcelik, Furkan Şahinuç, Fazli Can
Misinformation spread in online social networks is an urgent-to-solve problem having harmful consequences that threaten human health, public safety, economics, and so on.
no code implementations • 19 Apr 2022 • Cagri Toraman, Eyup Halit Yilmaz, Furkan Şahinuç, Oguzhan Ozcelik
Furthermore, we find that increasing the vocabulary size improves the performance of Morphological and Word-level tokenizers more than that of de facto tokenizers.
1 code implementation • LREC 2022 • Cagri Toraman, Furkan Şahinuç, Eyup Halit Yilmaz
The experimental results supported by statistical tests show that Transformer-based language models outperform conventional bag-of-words and neural models by at least 5% in English and 10% in Turkish for large-scale hate speech detection.
no code implementations • 2 Sep 2021 • Eyup Halit Yilmaz, Cagri Toraman
To provide additional information regarding the query and enhance the performance of intent detection, we propose a method for semantic expansion of spoken queries, called ConQX, which utilizes the text generation ability of an auto-regressive language model, GPT-2.