Natural Language Processing Working Together with Arabic and Islamic Studies

Summary

This is an interdisciplinary project which addresses the ICT call of "working together" by aligning ICT expertise and research interests from Computational and Corpus Linguistics, with Humanities research streams in Arabic and Islamic Studies, focusing on the Qur'an as a core text. It is also an international collaboration between the Universities of Leeds and Jordan, and further addresses the "working together" call via incoming and outgoing mobility in the form of Visiting Researcher placements in the School of Computing at Leeds (incoming) and the Centre for the Study of Islam in the Contemporary World at Jordan (outgoing). This agreement is proactive and novel, and has high impact, ensuring knowledge transfer from different methodological perspectives and cultures.

The study of Tajwid or Qur'anic recitation is a sub-field and taught module* in Islamic Studies programmes at both universities and elsewhere, and the original insight informing this project is to view Tajwid mark-up in the Qur'an as additional text-based data for computational analysis. This mark-up is already incorporated into Qur'anic Arabic script, and identifies prosodic-syntactic phrase boundaries of different strengths, plus gradations of prosodic and semantic salience through colour-coded highlighting of pitch accented syllables, and hence prosodically and semantically salient words.

The Computational Linguistics Module in Year 1 entails development and evaluation of software for generating a phonetically-transcribed, stressed and syllabified version of the entire text of the Qur'an, using the International Phonetic Alphabet (IPA). This canonical pronunciation tier for Classical Arabic will be informed and evaluated by Arabic linguists, Tajwid scholars, and phoneticians, and published in an updated version of the open-source Boundary-Annotated Qur'an Corpus [1], [2], preferably for LREC2 2014. The software will also be re-usable for Natural Language Engineering applications for Modern Standard Arabic, and for constructing dictionaries for Arabic language learners.

The Text Analytics Module in Year 2 implements statistical techniques such as keyword extraction3 to explore semiotic relationships between sound and meaning in the Qur'an, invoking a Saussurean-type view of the sign as '...a bi-unity of expression and content...' [5]. Our investigation entails: (i) text data mining for statistically significant phonemes, syllables, words, and correlates of rhythmic juncture [6], [7]; and (ii) interpretation of results from interdisciplinary perspectives: Corpus Linguistics (ICT); Tajwid science, plus Tafsir or Qur'anic exegesis (Islamic Studies); Arabic (Language and Literature); and Phonetics and Phonology (Linguistics).

In terms of ICT applications, the team will collaborate with stakeholders and beneficiaries to develop an associated or follow-on funding proposal for the UK Research Councils, to include publication of project software as an advanced corpus-query and visualization tool for Islamic Studies and Humanities scholars, plus Arabic language learners. This again represents an extension of the "working together" theme.

Finally, our approach is interdisciplinary and pioneers stylistic analysis of sound and rhythm encoded in writing as a semiotic system for religious and other literary texts. As such it is entirely novel and has direct implications for research-led teaching in both partner institutions plus a broad cross-section of research groups and user communities, namely: Natural Language Processing and Artificial Intelligence; Qur'anic and Islamic Studies; Arabic Language and Literature; Linguistics and Phonetics; Digital Humanities; and Psychology.

All references appear in Case for Support.

Impact Summary

Novelty and Originality: The original insight in this project [1] is to use Tajwid recitation markup of chunk boundaries in the Qur'an, plus other orthographical features, as untapped sources of text data for computational analysis. A principal objective is to establish quantifiable, linguistically rigorous links between the prosodic markup and the semantics of the Qur'an. This falls within exegetical science (Tafsir); it will make an original contribution to the Tafsir literature, which as yet provides no detailed analysis in this area, and be of interest to educated pious Muslims worldwide.

Debate and Controversy: Another objective is to develop a Tajwid-IPA and this might be seen as controversial since the Muslim mainstream has traditionally not treated any Qur'anic transliteration as the Qur'an proper. Any suggestion of a movement away from Arabic script to another language system would be seen as unacceptable by the mainstream - despite the advantages of doing so - and so this would almost certainly open up a debate, placing at least traditionalists and modernists in opposition to one another.

Sustainability: The project delivers re-usable algorithms for phoneticizing Arabic script for use in Arabic speech and language applications and lexicography. The Boundary-Annotated Qur'an Corpus [1] with canonical pronunciation tier will be the largest resource of its kind for training Arabic statistical language models and for Arabic linguistics and Islamic Studies. Re-usable algorithms for quantitative analysis of implicit prosody [3] in text will inform research on the link between text and spoken form in other languages. New interdisciplinary techniques for exploring religious and literary texts will inform AI research on deep text analytics. For Islamic Studies this research will open up the possibility of developing a new Qur'anic orthography comfortably rooted within modern phonetics, thus bringing an old oral tradition into a dynamic academic area of study. Current editions of the colour-coded Tajwid Qur'an are still inadequate in terms of providing the full gamut of Tajwid-phonetic values. This research is a stepping-stone towards a new Tajwid-IPA which we envisage would be welcomed by Arab linguists (though resisted by traditionalists). In this respect, we believe it to be a path-breaking research project.

Beneficiaries and how they will benefit: Beneficiaries span a broad cross-section of research groupings and user communities.

Science and Engineering: AI/NLP researchers who want a rich, gold-standard dataset for machine learning; language technologists for re-usable software to incorporate into systems; computational and corpus linguists as developers and users of annotated corpora and associated software; psycholinguists for conceptual models of prosody and language as a semiotic system.

Humanities: Richly annotated corpora are tools of the trade for Corpus linguists, Arabic scholars and researchers in Applied Linguistics, and can be exploited by lexicographers for dictionary construction; online Qur'anic resources have been identified as a priority by groups such as the Muslims in Britain Research Network and the UK Islamic Studies Network [4].

Economic: re-usable software for generating IPA transcriptions of Arabic words has market potential for publishers of dictionaries and language learning materials; the project aims to exploit re-usable software and annotated online resources for an industry-standard corpus-query and visualization tool.

Societal: The research will generate interest in the Middle East particularly in key centres of Islamic scholarship and is sustainable beyond the scope of this project, since a principal long-term objective is to provide a complete Tajwid markup system which can be used by institutions of Islamic learning in Britain - and worldwide.

References in Case for Support.