Prompt-Based Approach for Czech Sentiment Analysis

Proceedings of Recent Advances in Natural Language Processing, pages 98–108 Varna, Sep 4–6, 2023 https://doi.org/10.26615/978-954-452-092-2_011 98 ConfliBERT-Arabic: A Pre-trained Arabic Language Model for Politics, Conflicts and Violence Sultan Alsarra1, Luay Abdeljaber1, Wooseong Yang1, Niamat Zawad1, Latifur Khan1, Patrick T. Brandt1, Javier Osorio2, Vito J. D’Orazio3 1The University of Texas at Dallas, 2The University of Arizona, 3West Virginia University {sultan.alsarra,luay.abdeljaber,wooseong.yang, niamat.zawad,lkhan,pbrandt}@utdallas.edu, josorio1@arizona.edu, vito.dorazio@mail.wvu.edu Abstract This study investigates the use of Natural Lan- guage Processing (NLP) methods to analyze politics, conflicts and violence in the Mid- dle East using domain-specific pre-trained lan- guage models. We introduce Arabic text and present ConfliBERT-Arabic, a pre-trained lan- guage models that can efficiently analyze po- litical, conflict and violence-related texts. Our technique hones a pre-trained model using a corpus of Arabic texts about regional politics and conflicts. Performance of our models is compared to baseline BERT models. Our find- ings show that the performance of NLP models for Middle Eastern politics and conflict analy- sis are enhanced by the use of domain-specific pre-trained local language models. This study offers political and conflict analysts, including policymakers, scholars, and practitioners new approaches and tools for deciphering the intri- cate dynamics of local politics and conflicts directly in Arabic. 1 Introduction In the Middle East, political upheaval and carnage have long been issues (Blankenship, 2020). Deep divisions, geopolitical rivalries, and foreign med- dling have historically riven the area, from the Israeli-Palestinian conflict to the ongoing civil war in Syria. Even if the root causes of these conflicts are complex and multidimensional, the role that language and communication play in shaping the narratives that underlay them cannot be ignored. Language is commonly used as a strategy to rally support, defend violence, and discredit opposing viewpoints. Therefore, it is essential to develop effective methods for understanding and analyzing the role that language and texts plays in Middle Eastern politics and conflicts via news reports and other sources. Natural Language Processing (NLP) approaches can evaluate large amounts of text and have shown great promise in identifying patterns and insights that would otherwise be difficult to spot. Recent, pre-trained language models (PLM), like BERT (Devlin et al., 2018), have improved in efficiency for a range of NLP tasks, including sen- timent analysis, text categorization, and language synthesis. PLMs have received a lot of attention in the literature, but most of it has focused on English or other widely spoken languages; very few studies have examined how well they apply to Arabic. The Arabic language has a rich and complicated mor- phology, which has increased the requirement for highly advanced NLP tools that can meet the lan- guage’s expanding needs across a variety of fields and applications (Ameur et al., 2020). This research fills this vacuum in the literature by investigating the application of Arabic-specific PLMs for politics, conflicts and violence in the Middle East. We reference a variety of pertinent academic works, such as investigations into the na- ture of political violence (Asal et al., 2020), the function of language in conflicts (Webber …

Content truncated to 500 words