Learning causality for Arabic - proclitics

Sadek, J; Meziane, F

Abstract

The use of prefixed particles is a prevalent linguistic form to express causation in Arabic Language. However, such particles are complicated and highly ambiguous as they imply different meanings according to their position in the text. This ambiguity emphasizes the high demand for a large-scale annotated corpus that contains instances of these particles. In this paper, we present the process of building our corpus, which includes a collection of annotated sentences each containing an instance of a candidate causal particle. We use the corpus to construct and optimize predictive models for the task of causation recognition. The performance of the best models is significantly better than the baselines. Arabic is a less-resourced language and we hope this work would help in building better Information Extraction systems.

Citation

Sadek, J., & Meziane, F. (2018). Learning causality for Arabic - proclitics. Procedia Computer Science, 142, 141-149. https://doi.org/10.1016/j.procs.2018.10.469

• Journal Article Type: Article
• Acceptance Date: Oct 27, 2018
• Online Publication Date: Nov 15, 2018
• Publication Date: Nov 15, 2018
• Deposit Date: Nov 2, 2018
• Publicly Available Date: Nov 15, 2018
• Journal: Procedia Computer Science
• Print ISSN: 1877-0509
• Publisher: Elsevier
• Volume: 142
• Pages: 141-149
• DOI: https://doi.org/10.1016/j.procs.2018.10.469
• Publisher URL: https://doi.org/10.1016/j.procs.2018.10.469
• Related Public URLs: https://www.journals.elsevier.com/procedia-computer-science
• Additional Information: Access Information : This paper will be available open access under a CC-BY-NC-ND 4.0 licence once published in the journal.