machine learning and quran - the meccan and medinan verses
TRANSCRIPT
Is Machine Learning useful for Quranic Studies?
Abdul Baqi M. Sharaf
manual counting! excerpt from Ibn Kathir introduction
2
manual investigation! excerpt from badae' al-Fawaed by Ibn al-Qayyem
3
Quranic Resources
tanzil.net
• highly verified precise Quran text
• many translations:
http://tanzil.net/trans/
http://corpus.quran.co
m/
annotated linguistic resource which
shows the Arabic grammar, syntax
and morphology for each word in
the Holy Quran
Classification of
Surahs into Makki and
Madani
Makki and Madani * Debatable cases
8
Features [1/2]
verse of "sajdah" (prostration)
the aversion letter "kalla"
the phrase "Ya Ayyuha al-Naas"
the phrase "Ya Ayyuha
alladhina "
surahs with initials
story of Adam and Iblis
stories of previous nations
9
Features [2/2]
linguistic tools for denunciation,
excoriation, emphasis and oath
length of ayaat
eschatological topics
Jihad and battles of Islam
rulings on marriage and nursing
dialogue with ahl al-Kitab
pillars of Islam
10
Reduction to
computation
representation
Verses of Sajdah (K)-> search for
root “Sjd” -<< Noise: “al-masjid al-
Haraam” in Surah al-Baqarah
Quranic Initials (K) -<< Noise:
Surah al-Baqarah, ar-Raad
eschatological topics (K)-> Search
for: roots “Jannah”, “Jahannam”,
etc. <<- Noise: many occurance in
Surah Baqarah (D)
Jihad and battles of Islam (D) ->
Search for “jhd”, “qtl” <<- Noise: “
wa aqsamu billahe Jahda
aymanihim..” (Nahl)
11
Weka Representation
12
Decision Trees13
Decision Trees14
Top: random tree
Bottom: J48
Classifier outcome 15
13 – ar-Ra’d
55 – ar-Rahman
76 – al-Insaan
98 – al-Bayyenah
99 – al-Zalzalah
110 – an-Nasr
Clustering
16
Identify problem
• Suitable for machine learning
Identify feature sets
• Domain knowledge
• Existing books
Reduce to machine learning format
• Be aware of noise and inaccuracy
Run experiments
• Training set vs. Test set
Interpret results
• Verify and validate by domain experts
Machine Learning
Journey