10 Way Forward
There are voluminous forms of knowledge that had been passed down by the Islamic scholars, under the banner of Islamic sciences. Many of this knowledge had been cast as fixed and no new knowledge is possible; such is the case of closing the door of Ijtihad (independent reasoning)127. Similarly, we may ask, is there any new knowledge in the sciences of exegesis of Al-Quran, or the sciences of Al-Hadith?
Here we would like to distinguish between “new knowledge” and “new tools” for acquiring knowledge. The advancement of artificial intelligence (AI) has certainly received much attention of late. AI plays a major role in transforming Big Data into information, and from information to knowledge. Knowledge is where intelligence lies. We agree that machines can have intelligence. However, for machines to improve beyond intelligence towards wisdom, would be an impossibility. AI is an improvement in tools, through which we can improve our knowledge acquisition. But in the process, knowledge is for the benefit of humans, rather than machines.
Tools provide us baseline knowledge, that is sometimes used to confirm what we already know. For example, it is widely accepted that Allah is the central theme or subject matter of Al-Quran. The tools we used for Quran Analytics give clearer proof of that and, interestingly, a new perspective in the form of some unique characteristics of word networks. In this example, we say that the tools are providing affirmation to our knowledge and some new perspectives; but it does not add new knowledge. But tools can further be used beyond just existing known knowledge, such as to expand the analysis beyond its baseline. For example, we can further study why the name “Allah” is used in certain ways in the Al-Quran, through a bi-gram analysis. New insights from the analysis may yield new knowledge, which has not been discovered before. This is how tools are useful for exploring new knowledge.
Furthermore, we know that knowledge is dynamic. An example is the progress of knowledge through time. More discoveries are made as time passed. All that we want is to make sure that all knowledge is updated. AI presents the best modern tools to analyze data, because machines can scan, study, and look upon any possible large source of data, information, and knowledge, with fewer errors or fatigue (as we humans do). As an example, it is possible with AI tools today to study the differences between Ibnu Sina and Ibnu Rushd, who lived geographically far apart and centuries apart in time. Furthermore, the differences between them can be studied within the context of current-day knowledge and scholars.
In general, we say that AI tools help us to reaffirm and refine the “old existing knowledge” of Islam that we know, whether they are in the forms of traditions, fatwas, and interpretations. Furthermore, new tools can further strengthen the knowledge base. This is achieved by deployment of the new tools on existing old knowledge and re-represent them using the new tools. Examples of this include new visualization techniques, presentation in statistical forms or analysis, and summarization of knowledge in an abstract manner. Certainly, using new tools like text analytics on existing “data” like the Quran should reveal new knowledge and understanding about the Quran. It is also proof of the eternal relevance of the Quran as a book of knowledge.
New tools for studying Al-Quran
Studying Al-Quran is an important subject for Muslims and non-Muslims alike. In this book, we have been exploring the idea of mooting a specialized area called “Quran Analytics”. In various chapters, we have repeated the exercise of using NLP tools for different applications on the English translations of Al-Quran, in particular Saheeh International and Yusuf Ali. Throughout the various chapters, we keep highlighting various affirmations of knowledge of Al-Quran, which is already accepted as common knowledge. The concept of “Basheeran” (glad tidings) and “Nazeeran” (warnings) for the case of sentiment analysis is an example.
The other emphasis that we have been making is to highlight the approach of using the tools in search of discoveries. For this purpose, we introduced tools that are commonly used in NLP exercises, such as word collocations, word networks, and word modelings. Since the focus is on the translated Al-Quran, the analysis used is tied to the English structure of the translations. Despite these limitations, many interesting insights were obtained. As an example, we could detect the resiliency of the messages in Al-Quran, regardless of the translation method used. Despite differences in time, Yusuf Ali and Saheeh contain similarities in many structures that are attributed mainly to the original source of the texts, which is the Arabic Al-Quran.
What we have attempted represents only some sampling of analyses with the application of new tools in NLP applied on the English translations of Al-Quran. We have not even attempted to deal with the Quranic Arabic language, which is another vast area by itself. It is known that the Quranic Arabic language influences the Classical Arabic language which in turn was the base for most of the works in Islamic knowledge. The language is not the creation of Muslims, but rather was passed down from the Arabs, making it the lingua-franca of Islam and Muslims. The Quranic language dominated the scientific and knowledge dissemination of the world for over a thousand years. All knowledge and scientific works, except for literary pieces (particularly among the Persians) were in the Arabic language. The sciences of the Arabic language (a major issue in Computational Linguistics and Natural Language Processing today), evolved into well-established sciences very early on in the history of the language. This is evident from the early works of Al-Farahidi, followed by his followers, among them is Al-Sibawayhi. What they had accomplished is to establish the rules of grammar for the Arabic language.
Among the task of CL and NLP nowadays is to convert the rules of language into a computational model. There no better sources that we can rely upon to develop the models (for Classical Arabic) than these classical works. By right, these works (and works of many scholars in later periods), present a gold mine of data and process, which can be converted into concise computational models of the language.
A full-scale work involving the Quranic Arabic language, the various sources for the knowledge on Al-Quran (or Ulum Al-Quran), deploying various tools of CL and NLP, and development of models and applications for knowledge of Al-Quran, is what we termed as “Quran Analytics”. This is what we envisaged to be developed in its full capacity.
Limitations
Throughout the book we have demonstrated many applications of CL and NLP on Al-Quran, using mainly the selected English translations of Saheeh and Yusuf Ali. Admittedly, the selection and focus are narrow and limited. The idea is to demonstrate some of the tools using the R programming language. Despite these limitations, we need to emphasize that what we have demonstrated is the ability of the tools to expand the knowledge further.
Our focus in this book is mainly on applying the tools and modeling, but not on the meaning and interpretations. Neither do we have a focus on the aspect of the language of Al-Quran. However, given a domain knowledge on a particular area, like a method of exegesis of Al-Quran such as by Imam Ibnu Katheer, we can quickly expand the subject with ease as shown in Chapter 9.
The ego network of verses from Tafseer Ibnu Katheer reveals many of his understanding of Al-Quran which is not as evident if we read the works verbatim. The only way we can see this is through the network representations. Even then, our work so far is to record and display, without proposing deeper meanings and interpretations of the arrangements. We purposely wanted to avoid the interpretations since it requires deeper knowledge of the subject under discussion.
In any case, we would say that the limitations of our works are only by choice rather than by default. Any interested student or researcher can take the direction which we started with and make many other findings and learnings from the old knowledge as well as discoveries of new knowledge.
Direction of future works
It is our belief, based on the findings in this book, that the direction of future research for Quran Analytics is tremendous and vast. This is encouraged by two facts. First, the CL and NLP tools are continually improving which will make many more tools and methods available for applications on Quran Analytics. Second, the amount of digitized data becoming available is ever increasing which allows us to perform more analysis based on classical sources of knowledge as well as contemporary sources. All are now becoming more easily available.
The works however require caution, since implementations of modern tools are not value-free. For example, sentiment analysis based on English texts assume certain value judgments which may not fit the Quranic language. Therefore, there is also a real need of developing models from scratch, instead of relying on ready-made tools.
The research approach has been towards “open and reproducible” research. What is meant is basically that research is encouraged to use openly sourced data, openly sourced codes (or programming languages), and should be available for inspection and reproduction by others. We fully agree and adhere to this approach, whereby all the codes of our works are openly available on Github, written mainly using R programming language as an open-source tool.
We also follow closely the works of development of corpora for the Islamic texts and encourage development along those lines. We also would like to invite more people, especially those with domain knowledge in Islamic sciences to adopt a similar approach of using AI tools for Islamic sciences applications.
Concluding remarks
Our sincere hope is to spur more works and usage of AI tools for the furtherance of the development of Islamic knowledge. Quran Analytics, in our view, is the starting point, and this book is an introduction to the subject. There are many lessons and improvements possible using modern AI tools for Islamic knowledge.
For this field to expand, we believe that it is important for people from various disciplines and backgrounds to collaborate. It requires people with a background in computing and AI, with people having knowledge of Islamic sciences, and knowledge of Classical Arabic language, combined. This inter-disciplinary effort is a must if progress and improvements are desired.
We call upon everyone to join us in these efforts, and we pray that Allah will reward all of us in this endeavor.