Tonirodriguez at CheckThat!2024: Is it Possible to Use Zero-Shot Cross-Lingual Methods for Subjectivity Detection in Low-Resources Languages?

Subjectivity detection is a key task within natural language processing due to the challenges generated by new forms of journalism, the proliferation of misinformation and fake news, and existing concerns about the quality and integrity of journalism. Although subjectivity detection is an existing challenge in all languages, the amount of resources available to build these types of applications varies greatly among languages. In this paper, we present our participation in the CLEF2024 CheckThat! Lab Task2 [1], where we have attempted to apply Zero-Shot Cross-Lingual transfer techniques using the datasets for the five languages provided in Task2 (English, German, Italian, Bulgarian, and Arabic). For this, we have fine-tuned two multilingual models, mDeBERTa v3 and XLM-RoBERTa, on a subset of the dataset consisting of three of the languages provided in Task2, specifically English, German, and Italian, and we have applied Zero-Shot Cross-Lingual transfer to the other two languages available in Task2, Arabic and Bulgarian.