Hi Kais and everybody else,
I want to point out that the questions you suggested all revolve around one
word:
Allah - Who is Allah?
Muhammad - Who was Muhammad?
Islam - What is Islam?
prophets - Which prophets are mentioned in the Quran?
Solomon - Is Solomon mentioned in the Quran?
Jesus - What does the Quran say about Jesus?
religions - List the religions mentioned in the Quran.
If you stick with questions like these, what is the difference with an index or
concordance? Or is your intention to generate a coherent answer from the
relevant verses? If so, I would say that paraphrasing the word of God is not a
good idea (this objection is similar to my objection to the pronoun
"resolution"). Also be aware that much of the Quran is context dependent and so
one cannot lift one verse up and present it out of its context; important
nuances might very well be lost (not to mention the extensive literature on
asbab al-nuzul, tafsir, sirah, ahadith etc. to contextualize the Quran even
more).
Yes, we live in a time where it seems that everything has to be summarized in
one catchy quote (i.e. if it doesn't fit in a Twitter tweet, it's not important
to read) but I honestly don't think an approach like that to the Quran is a
correct one. The very character of the Quran - and I mean for example its lack
of one big narrative and its apparent contradiction throughout the text -
forces the reader to spend time on it, to read bigger parts of it than just
some verses here and there. To decompose the Quran to its verse-structure (or
do you want to cut into even smaller pieces for this?) and put it back together
in an order you deem best is a guarantee for losing its aesthetics and thereby
its unique character.
So to conclude, is what you want pretty much the same as an index/concordance?
Then go with that. If you think the QA-system would be different from that,
please take note of the pitfalls I've indicated here.
Just my 2 cents of course.
Eric van Lit
M.A. student McGill University
> Date: Sun, 28 Feb 2010 10:48:23 +0000
> Subject: Re: Advice Required - Dialog System for the Quran
> From: [email protected]
> To: [email protected]
>
> Hello,
>
> This discussion thread is getting quite interesting indeed. What do
> you all think of START - the natural language question answering
> system?
>
> http://start.csail.mit.edu
>
> I think it's pretty good for simple facts, especially as it usually
> cites the original source resource for each answer. With regards to a
> Quranic QA system, I think that so far the general consensus on this
> mailing list is that there could be two types of system:
>
> (1) A very basic fact system, which can answer simple questions from
> knowledge in the Quran, backed by Quranic verses.
>
> (2) A much more advanced system, which combines the Hadith, Quran, and
> other parts of the Islamic faith to give fatwa-style answers using
> advanced inference and automated deduction.
>
> Well, we have to start somewhere ... so I'm keen to focus on system
> (1). The question for me is, would a simple fact system still be
> useful, without advanced inference? From a practical perspective, I
> believe that it would be, although of course not as useful as system
> (2). A very-simple QA system might be useful for students, younger web
> users, and the general public who might be interested in simple
> questions about Islam. Probably not so useful for those already
> familiar with the Quran and the religion in more detail and are after
> finer clarification. From an Academic perspective, system (1) is also
> very useful because:
>
> - It might be one of the first few attempts at a Quran QA-system
> (would give a baseline for future research in this area).
> - It would be interesting to know how the genre of the Quran affects
> QA in general, i.e. is there anything special about this domain? Is QA
> for the Quran the same or any different from general QA?
> - Given that the Quran is highly studied (and now also highly
> annotated linguistically) does this extra information help beyond
> basic keyword search?
> - This simple system is probably a step in the right direction for a
> more advanced system (2)
>
> Based on recent discussion, I would suggest that a good starting point
> would be a simple QA fact system. Note that "simple" here refers to
> the system from a user's perspective. From the perspective of those
> designing and building such a system, things which look "simple" to a
> user can be very hard indeed in the field of AI :-), if not sometimes
> nearly impossible!
>
> Would anybody be interested in helping to compile a list of basic fact
> questions about the Quran, backed by specific verses? Here are some
> simple fact-based questions that come to mind:
>
> - Who is Allah?
> - Who was Muhammad?
> - What is Islam?
> - Which prophets are mentioned in the Quran?
> - Is Solomon mentioned in the Quran?
> - What does the Quran say about Jesus?
> - List the religions mentioned in the Quran.
>
> (Inspired by http://start.csail.mit.edu)
>
> The good thing is that once we get a basic system up online, we can
> start to monitor the system's log files to see which questions are
> most frequently asked (we would expect a Zipfian distribution) and we
> can later tune the system to give good responses to the most common
> questions. Hopefully each of these questions could be answered via an
> extract of a relevant verse from the Quran (within reason). Ideally, I
> would like to compile a list of hundreds (or even possibly thousands)
> of questions with simple answers from verses of the Quran.
>
> After that, there could be lots of other possibilities. Given a large
> list of question and answers, this could be the final knowledge base,
> and we map new questions to existing answers. Otherwise, given a large
> list, there may be a reasonable way to think about automating the
> process of question/answer extraction to a reasonable level of
> accuracy - so that future compilation of the knowledge base might be a
> case of manual verification of automated suggestions.
>
> I'm looking forward to any suggestions on the first stage of this new
> project: constructing a large list of simple questions and answers
> relating to verses of the Quran.
>
> Kind Regards,
>
> -- Kais
>
>
> > On Sun, Feb 28, 2010 at 8:36 AM, Abdul-Baqi Sharaf
> > <[email protected]> wrote:
> > Dear All;
> >
> > I was following this thread with interest. Actually I am not very much
> > optimistic about trusted automated QA system on the Qur'an or Islam that
> > would assume the position of a Qur'anic scholar. However, a simple factoid
> > questions to someone who does not know much facts about Qur'an/Islam can be
> > produced.
> >
> > I think automated QA is the greatest challenge for artificial intelligence
> > and Alan Turing's challenge still remains a challenge for computational
> > community. Shaping this challenge for the quranic context would mean, that
> > when a user asks a quranic questions he/she would not be able to tell who is
> > answering him/her from behind the screen: a qur'anic scholar or an automated
> > dialogue system :-) so, seconding El-Haj I also think any system should not
> > assume the position of a Mufti.
> >
> > Once I gathered few questions from tafsir section of islam-qa.com and it
> > seemed to me that majority of these questions are geared towards general
> > Muslims who has fair knowledge about the Qur'an but wants clarification on
> > apparent contradictions or confusion about two verses. I would imagine that
> > some good progress could be achieved if the Qur'anic is annotated with much
> > more extra-linguistic tags including information taken from books of tafsir,
> > hadith, fiqh, etc and then sophisticated inference rules are applied, but
> > then again any answer should be first checked by a Qur'anic scholar before
> > passing to the user which then defeats the purpose of a dialogue system
> > which assumes instant answers.
> >
> > However, I think computational linguistics and text mining can be of great
> > help to a qur'anic scholar who can filter any automated output and accept
> > the ones that is correct.
> >
> > Having said that, I am not discouraging this idea of a dialogue system, but
> > rather requesting to define the project properly in terms of the type of
> > questions this system will attempt to answer instantly.
> >
> > I will be interested in collecting a set of factoid question-answers if this
> > projects goes forward.
> >
> > best wishes,
> > Abdul-Baquee
> >
> > On Sun, Feb 28, 2010 at 3:56 AM, El-Haj, Mahmoud <[email protected]> wrote:
> >>
> >> Dear All,
> >>
> >> Very impressive indeed.
> >> I just have the following questions, concerns and suggestions regarding
> >> the Quranic Question Answering System.
> >>
> >> Building a Quranic Question Answering system is great but you need to
> >> focus on what type of questions are you going to ask!
> >> Is it questions on Belief, juristic, transactions, ...etc.
> >>
> >> Trustworthy Resources! What does it exactly mean? Are you intending to
> >> just copy/paste questions and answers from different websites? Which sounds
> >> retrieving process only.
> >> What I mean is training a Quranic QA system with Quranic resources is
> >> different from training a general QA system using a training data, as the
> >> domain is huge and the testing data is different from the training one
