European Journal of Scientific Research ISSN 1450-216X Vol.37 No.1 (2009), pp.36-40 EuroJournals Publishing, Inc. 2009 http://www.eurojournals.com/ejsr.

htm

Dialogue-based Visualization for Quranic Text

Aida Mustapha Faculty of Computer Science and Information Technology University Putra Malaysia, 43400 UPM Serdang, Malaysia E-mail: aida@fsktm.upm.edu.my Tel: +603-8946-6554; Fax: +603-8946-6577 Abstract Visualization is a human cognitive activity and is not something that a computer does. A good visualization for Quranic text enables us to communicate the content with clarity, precision, and efficiency. Being a large information source, a visualization system for Quranic text is required. This paper proposes AQILAH, dialogue-based information visualization for Quranic text. AQILAH facilitates the interaction and rearrange the content based on need-to-know basis, as opposed to sequential order as in recitals.

Keywords: Dialogue Systems, Information Visualization, Quranic Text

1. Introduction
Information visualization is the formation of mental visual images, the act or process of interpreting in visual terms. Information visualization combines aspects of imaging, graphics, scientific visualization, human-computer and human-information interactions, as well as information technology (Robertson, Card, and Mackinlay, 1993). Basically it is a mapping from data representation to perceptual representations that maximize human understanding. In knowledge management, visualization is a type of internalization process (Marwick, 2001) that transformed the knowledge from explicit to tacit. Explicit knowledge in Quranic texts are the verses (ayat), while the tacit knowledge is the mental understanding based on the users experience exploring the Quranic text. The Quranic text consists of 114 chapters (surah) of varying lengths. Each surah also differs from one another in terms of the number of verses (ayat) (Shakir, 1983). Previous works on Quranic text mainly focus on retrieval techniques through class menus, topical and keywords, and stemming and block-level link (Atiyah, 1996; Aljlay and Frieder, 2002; Noordin and Othman, 2006), while visualization techniques focus on terms visualization (Ismail et al, 2007) using graphical or imaging techniques like hyperbolic trees and dynamic queries, which are hierarchical or diagrammatic in nature (Hamzah, 2007). The main argument behind graphical representation is that visual format assists users in interpreting the data or information in the form of understanding or mental images. Nonetheless, there are two kinds of statements in verses of Quranic text. First are statements that provide a clear, concise indication of the idea underlying a particular passage or passages (muhkamat). For instance, verses those refer to the creation of man from out of a drop of sperm deliver the literal meaning of biological origin of human beings. Second are allegorical or symbolic statements (mutashabihat) that are not entirely clear as they convey several meanings, while some carry other literal indications, although such indication might not be intended. For instance, verses those refer to Gods being such that He is being indefinable and infinite in time and space are far from

Dialogue-based Visualization for Quranic Text

37

describing Him, but are only to imply what He is not: namely, not limited in either time or space, not definable in terms of comparison, and not to be comprised within any category of human thought. Because mutashabihat verses are being conveyed as generalized metaphors that are well beyond the comprehension of human perception, mapping the data representations in the form of Quranic terms should be represented differently from visual images. This paper proposes a dialoguebased visualization system for Quranic text, whereby knowledge queries are in the form of natural language, while the responses drawn from the knowledge source of Quranic text are also in natural language. The remainder of this paper introduces the dialogue-based Quranic visualization, the proposed architecture, the prototype implementation of the visualization system, followed by conclusions and some direction to future works.

2. Dialogue-based Quranic Visualization

The main idea behind Quranic visualization is the huge amount of data encoded in the body text; hence an independent verse (ayat) on its own is virtually incomprehensible. Quranic verses are to be interpreted as a whole, within the context of the chapters (surah) in order to maximize the understanding. Because interpretation of Quranic texts is highly contextual according to the chapters, a visualization system must be able to capture co-occurrences of concepts across the chapters in its original textual form, rather than in graphical or diagrammatical forms. For instance, repetitive references to the origin of human being can be quoted in many varying formulations like Was he not a drop of sperm emitted (In lowly form) in [75.37] and Verily We created man from a drop of mingled sperm in [76:2]. Although question-answering or information retrieval systems are able to elicit a specific reference in answer to a query, the nature of knowledge organization in Quranic texts made the user learning model confined to the scope of the query. Dialogue-based systems, on the other hand, respond to user queries in a more elaborated fashion through series of verses related to the queries. Hence, a dialogue-based visualization is more akin to formation of mental model as captured through conversation with a human preacher. A dialogue-based visualization system advances the internalization process through construction and reconstruction of user mental models until all knowledge and relationships within the Quranic text is perceived by the users in its simplest form, which is in natural language. A hypothetical scenario is when a PhD student is asked on the research he is working on. The explanation would start from a single broad answer (i.e., Artificial Intelligence) and only further questions will lead to more detailed answers (i.e., to improve the efficiency of algorithm X). It is very rare, unless the opposing conversation participant is working in the same field, that a direct answer is obtained. Basing on the same ground, when a human preacher is asked on the biological origin of human, it is very rare to receive a response in the form of specific reference to a particular verse, but rather in the form of big conceptual picture with reference to many verses. In the essence, conversation dialogues encourage elaboration of concepts in a series of utterances rather than in a form of hierarchical relationships.

3. Prototype of Aqilah
The proposed dialogue-based visualization system is implemented as a prototype called AQILAH, based on the learnable conversational agent framework (Mustapha and Abu Bakar, 2005). Since the architecture is reusable, the focus in this prototype development is the knowledge base preparation. Nonetheless, the preparation must adhere to two important characteristics in dialogue conversations. One, the basic unit in a dialogue is in the form of sentence or utterance. Two, dialogue utterances are produced in turns by two speakers; hence one speaker may produce more than one utterance. Figure 1 shows the general architecture in AQILAH.

38
Figure 1: The Architecture of AQILAH

Aida Mustapha

User Task Level Learnable Conversational Agent Individual Conversation Log

Natural Language Parser Knowledge

Learning Engine

Knowledge Database

Communication Random Response

Associate Response

Words Dictionary

The architecture in AQILAH represents the inevitable dependencies and relationships between the various knowledge base components in the system, which facilitates user understanding of the internal interaction in the dialogue-based visualization system. The dotted lines indicate different interaction roles by the architecture (Mustapha and Abu Bakar, 2005). Based on Figure 1, AQILAH is equipped with a minimal technology of simple keyword-based parsing to parse text-based input of human natural language. AQILAH responds in natural language response based on the Quranic text by means of extracting keywords from an input string and performing keyword matching against the knowledge base, a dictionary of associative words, and a backup of random answers. Each priori response contains three wildcards per response and each wildcard may also leads to other keyword so as to increase the probability for a successful match. 3.1. Source of Knowledge Base Al-Quran is originally written in Arabic language, but the Quranic texts used as the source of knowledge in this prototype is the English translations (Shakir, 1983). The texts are organized into 114 chapters (surah) of various lengths with specific number of verses (ayat) in each chapter as shown in Table 1. In total, the Quranic text consists of 6,236 verses in 114 chapters.
Table 1: The number of verses (ayat) in the first ten chapters (surah)
Chapter Name The Opening The Cow The Family of Imran The Women The Dinner Table The Cattle The Elevated Places The Accessions Number of Verses 7 286 200 176 120 165 206 75

Chapter Number 1 2 3 4 5 6 7 8

Dialogue-based Visualization for Quranic Text

39

Due to the nature of dialogue utterances, the Quranic texts are segmented into singular verses that can be manipulated as individual dialogue utterances, which in turn, may be used on its own or in a series of utterances in case of multiple utterance in a turn. There are three knowledge bases involved in the process of preparing verses into response utterances, known as the Master, the Associate, and the Random knowledge base. The knowledge bases are organized in the form of input pattern and response templates. In the Master knowledge base, pattern is the keyword to match with user inputs and the response templates are in the form of references to verses. While patterns are indexed according to chapter numbers, singular utterances are indexed according to Quranic reference citation, for instance [8:73] refers to verse number 73 in chapter 8 of The Accessions. The wildcard ontology is designed as a flat file based on the required architecture (Mustapha and Abu Bakar, 2005). 3.2. Associative and Random Base The Associate knowledge base is required to store the associative lexicons, or words with similar meanings. However, the design is not the same with thesaurus or dictionaries, whereby the contents are meaning-bearing words in English (i.e., big vs. enormous). This associative base stores Islamic terms of similar meaning like an Arabic word nutfah, which literary means minute quantity of liquid but scientifically referred as sperms. The word of similar meaning serves as secondary knowledge base whose function is to replace the non-existing keyword with the similar words and return to search in the Master with the new keyword. The Random knowledge base, on the other hand contains random responses to return to user when search for patterns have been exhausted. The main idea is diversion by pointing to a new verse as a way to change current topic, hence admitting failure to match a response in a more natural way. It is particularly important to maintain a high probability of success in returning a response while keeping the conversation going.

4. Conclusions and Future Works

This paper began by showing that a dialogue system can be a good visualizing tool to navigating and learning Quranic text, with interaction mechanism being the closest to a human preacher. This is because Quranic text is a huge body of corpus and learning is usually not in sequential order as recitals. The information need is usually less specific and more general with hope to find the best explanation for the knowledge quest. We then presented AQILAH, and further detailed the knowledge base for AQILAH. Nonetheless, we would like to highlight some implications of using Quranic text as the knowledge source. One is that, the source of Quranic text comes in written style thus resulting in rigid dialogues utterances because the visualization system converses in the manner as provided by the source text. However, given the fact that the knowledge from Quranic text should be treated earnestly, the authoritative style of dialogue utterances is therefore suitable for this type of application. Second, the chapter-length surah is hard to be subdivided into series of turns as a single-utterance style in dialogue systems in effort to preserve the meaning and context. To resolve this, we adopted the Quranic text at face value. We used individual verse (ayat) as a single utterance, therefore allowing multiple references to a specific ayat because a particular ayat may be referred by different surah and it may also appear more than once in the same surah. This research is hoped to pursue in two directions. The first is to explore into a probabilistic corpus modeling based on the ground that most verses are repetitious in multiple places. We hope to capture the probabilities for each verse and in turns collate the duplications to be considered as the best return response during a particular session of so-called dialogue-based preaching. The next direction is to incorporate an ontology (Yusof et al, 2009) that focuses on the structure of content so mapping from

40

Aida Mustapha

the knowledge base to the pattern response templates can be improved. We hope to provide comprehensive source for the dialogue-based Quranic visualization system.

References
[1] Aljlayl, M. and Frieder, O., 2002. On Arabic Search: Improving the Retrieval Effectiveness via a Light Stemming Approach, In Proceedings of the Eleventh International Conference on Information and Knowledge Management, ACM Press, New York, pp. 340-347. Atiyah, H.M., 1996. Quranic Text: Toward a Retrieval System, In Proceedings of the IIIT, Virginia. Hamzah, R., 2007. Visualizing Al-Quran: The New Innovation of Reciting Al-Quran, In Proceedings of International Conference on Engineering Technology, Kuala Lumpur. Ismail, N.K., Abd Rahman, N., Abu Bakar, Z. and Tengku Sembok, T.M., 2007. Terms Visualization for Malay Translated Quran Documents, In Proceedings of the International Conference on Electrical Engineering and Informatics, Bandung, Indonesia. Marwick, A., 2001. Knowledge Management Technology, IBM Systems Journal 40:4. Mustapha, A. and Abu Bakar, A., 2005. LUQMAN: Learnable Conversational Agent for Home-based Education, In Proceedings of the International Conference on Information Technology and Multimedia, UNITEN, Malaysia. Noordin, M.F. and Othman, R., 2006. An Information Retrieval System for Quranic Text: A Proposed System Design, In Proceedings of Second Information and Communication Technologies, 1, pp.1704-1709. Robertson, G., Card, S., and Mackinlay, J., 1993. Information Visualization using 3D Interactive Animation, Communication of ACM 36(4):57-71. Shakir, M.H., 1983. The Holy Quran, Tahrike Tarsile Quran. Yusof, R.J.R, Zainudin, R., Baba, M.S. and Yusoff, Z.M., 2009. Visualization Systems Supporting the Reading of Arabic Document for non-Arabic Speakers, Information Technology Journal 8(1), 16-27.

[2] [3] [4]

[5] [6]

[7]

[8] [9] [10]