Constructing a Bilingual Hadith Corpus Using a Segmentation Tool

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 3390–3398 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 3390 Constructing a Bilingual Hadith Corpus Using a Segmentation Tool Shatha Altammami12, Eric Atwell2, Ammar Alsalka2 King Saud University1, University of Leeds2 Saudi Arabia, UK Shaltammami@ksu.edu.sa1 {Scshal, E.S.Atwell, M.A.Alsalka}@leeds.ac.uk2 Abstract This article describes the process of gathering and constructing a bilingual parallel corpus of Islamic Hadith, which is the set of narratives reporting different aspects of the prophet Muhammad’s life. The corpus data is gathered from the six canonical Hadith collections using a custom segmentation tool that automatically segments and annotates the two Hadith components with 92% accuracy. This Hadith segmenter minimises the costs of language resource creation and produces consistent results independently from previous knowledge and experiences that usually influence human annotators. The corpus includes more than 10M tokens and will be freely available via the LREC repository. Keywords: Hadith, parallel corpus, NLP, language resource, Arabic, English. 1. Introduction Current advancements in Artificial Intelligence (AI) and Natural Language Processing (NLP) led to at- tempts at computerising a range of tasks that require domain experts. One of the research areas that caught interest in AI methods is the study of religious texts to enhance understanding and discover new embedded knowledge. However, the main obstacle of such studies is the lack of annotated corpora suitable for religious-oriented text-mining tasks. In this work, we aim to enrich an under-resourced religious text, Islamic Hadith, which is the set of narratives reporting the words, actions and habits of the prophet Muhammad. Although Hadith’s importance is second to the Quran’s (the Muslims holy book), most laws and legislations are obtained from the Hadith due to its larger scope and incor- porated details. Yet, Islamic computational studies has focused on the Quran, leaving Hadith relatively unexplored. One possible reason is Hadith’s vast and varying literature with inconsistent structure that makes collecting them in a well-structured corpus a challenging task. Research in the area of Hadith computation is still in its infancy (Bounhas, 2019). Yet, there is an annual increase in the number of published papers indicating it is gaining wider attention from multi-disciplinary researchers (Azmi et al., 2019). In such work, re- searchers gather their own dataset from different sources and sometimes manually process them (Luthfi et al., 2018). This indicates the field is lacking adequate language resources and reusability is limited since the collected datasets are not published for use in other research projects. Hence, it is unfeasible to establish benchmarks, compare results or set evalu- ation measures (Guellil et al., 2019), which makes establishing a Hadith Common Dataset Initiative an imperative. Through this project, we initiate the Hadith Common Dataset by introducing a well-structured parallel corpus of Hadith in its original classical Arabic text and corresponding English translations obtained from well-known Hadith books. To the best of our knowledge, no parallel corpus of Hadith is freely available to the research community. The accessible data is scattered around …

Content truncated to 500 words