1
00:00:00,660 --> 00:00:02,220
In God's name, the Compassionate, the Merciful.

2
00:00:02,220 --> 00:00:03,920
Praise be to God, Lord of all worlds.

3
00:00:03,920 --> 00:00:07,000
Peace and blessings upon the noblest of God's
creation,

4
00:00:07,000 --> 00:00:11,840
our Prophet Muhammad, his family, companions, and
followers until the Day of Judgment.

5
00:00:12,260 --> 00:00:17,780
Today, God willing, I will discuss the Holy Quran
and

6
00:00:17,780 --> 00:00:23,340
the author's fingerprint: how we can use it to
distinguish

7
00:00:23,340 --> 00:00:29,890
between different writers. Let me begin by recalling
that when a messenger is sent,

8
00:00:30,320 --> 00:00:33,700
God, exalted be He, sends him with two things:

9
00:00:33,700 --> 00:00:38,220
a miracle showing that he is a messenger from God,

10
00:00:38,250 --> 00:00:41,910
and the teachings he must convey to God's servants.

11
00:00:42,070 --> 00:00:46,590
Moses had his staff,

12
00:00:46,590 --> 00:00:50,530
and the Torah. Jesus healed those born blind and
those with leprosy,

13
00:00:50,530 --> 00:00:52,610
and he also had the Gospel.

14
00:00:52,610 --> 00:00:58,730
For the Prophet, peace be upon him, the miracle and
the teachings are one and the same.

15
00:00:59,120 --> 00:01:00,960
The Holy Quran is the miracle.

16
00:01:03,180 --> 00:01:08,140
The Quran itself challenged the Arabs to produce a
surah like it.

17
00:01:08,320 --> 00:01:13,320
God says: 'If you doubt what We have revealed

18
00:01:13,320 --> 00:01:19,100
to Our servant, produce a surah like it, and call
your witnesses besides God,

19
00:01:19,100 --> 00:01:24,340
if you are truthful.' Another verse says: 'Or do
they say

20
00:01:24,340 --> 00:01:28,520
he invented it? Say: Then produce ten invented
surahs like it,

21
00:01:29,030 --> 00:01:32,310
and call whoever you can besides God, if you are
truthful.'

22
00:01:32,790 --> 00:01:38,010
Another verse says: 'If humans and jinn joined
together to produce

23
00:01:38,010 --> 00:01:42,990
the like of this Quran, they could not, even if they
helped one another.'

24
00:01:44,380 --> 00:01:49,260
Now let us move straight to the content of this
presentation.

25
00:01:49,260 --> 00:01:54,940
I'll briefly introduce the author's fingerprint and
how

26
00:01:54,940 --> 00:01:56,940
we can extract it from different texts.

27
00:01:57,580 --> 00:02:02,890
We can use it to group texts automatically

28
00:02:02,890 --> 00:02:08,290
by their author's fingerprint, or by the time

29
00:02:08,290 --> 00:02:10,540
or period in which they were written.

30
00:02:11,790 --> 00:02:16,270
After this introduction, I'll move to an experiment

31
00:02:16,270 --> 00:02:20,950
with prose: an experiment involving ten authors.

32
00:02:21,050 --> 00:02:25,990
Then I'll move to the experiment with the Holy Quran
and Arabic poetry.

33
00:02:27,090 --> 00:02:31,230
I'll finish with a conclusion and the findings.

34
00:02:32,260 --> 00:02:37,480
Stylometry, or computational stylistics, as we might
call it,

35
00:02:37,480 --> 00:02:41,840
is a technique we can use on a computer

36
00:02:41,870 --> 00:02:44,550
to identify an author's fingerprint.

37
00:02:44,550 --> 00:02:50,190
It can also identify the different periods or times

38
00:02:50,190 --> 00:02:54,270
when these works were written,

39
00:02:54,270 --> 00:02:57,010
whether poetry or prose.

40
00:02:58,630 --> 00:03:02,190
This algorithm uses hierarchical clustering.

41
00:03:02,350 --> 00:03:07,890
It extracts the most frequent words from the texts,

42
00:03:07,890 --> 00:03:13,570
because these are used in different kinds of
writing,

43
00:03:13,570 --> 00:03:15,130
regardless of the subject.

44
00:03:15,130 --> 00:03:20,430
The same writer uses the same prepositions,
conjunctions, and similar words

45
00:03:20,430 --> 00:03:22,650
when writing two different texts:

46
00:03:22,650 --> 00:03:24,630
one about economics and another about politics.

47
00:03:24,710 --> 00:03:27,330
These words form

48
00:03:27,850 --> 00:03:32,630
the mathematical vector that represents the author's
fingerprint.

49
00:03:33,010 --> 00:03:38,530
We then use established distance measures to group
the texts

50
00:03:38,530 --> 00:03:43,830
into clusters based on similarities in their writing

51
00:03:43,830 --> 00:03:49,410
style. These algorithms treat a text

52
00:03:49,410 --> 00:03:50,930
as a mathematical vector.

53
00:03:51,050 --> 00:03:56,150
For example, the blue vector in this diagram

54
00:03:56,150 --> 00:03:58,810
is represented in three dimensions.

55
00:04:03,180 --> 00:04:08,860
In the same way, we generate mathematical vectors
for all the texts,

56
00:04:08,860 --> 00:04:14,440
with 100 dimensions. These vectors come from the
most

57
00:04:14,440 --> 00:04:18,100
frequent words in the corpus: the collection of
texts being analyzed.

58
00:04:20,260 --> 00:04:24,920
For example, here are the vectors generated

59
00:04:24,920 --> 00:04:29,340
for the texts labeled H1 and onward.

60
00:04:36,900 --> 00:04:42,380
The algorithm measures the distances between the
generated vectors,

61
00:04:42,380 --> 00:04:44,560
and merges the closest ones.

62
00:04:44,720 --> 00:04:50,340
Then

63
00:04:50,340 --> 00:04:53,860
it merges entire groups according to similarity.

64
00:04:53,860 --> 00:04:56,520
A different or outlying text, such as J1,

65
00:04:56,520 --> 00:04:58,240
ends up in a different cluster.

66
00:05:00,200 --> 00:05:02,560
This is a very important experiment:

67
00:05:02,560 --> 00:05:04,880
the experiment with ten travelers.

68
00:05:05,040 --> 00:05:08,480
We used ten different books,

69
00:05:08,480 --> 00:05:12,340
by writers who lived at roughly the same time,

70
00:05:12,340 --> 00:05:15,400
and wrote about the same subject.

71
00:05:15,400 --> 00:05:20,440
We wanted to see whether we could use

72
00:05:20,440 --> 00:05:24,840
the most frequent words to distinguish between these
different writings.

73
00:05:25,000 --> 00:05:28,720
We took about ten samples from each book.

74
00:05:29,490 --> 00:05:33,130
The samples were selected randomly.

75
00:05:33,130 --> 00:05:34,930
So, from each book,

76
00:05:34,930 --> 00:05:36,850
we had passages from the start, the end,

77
00:05:36,850 --> 00:05:42,270
and the middle. Each sample was preprocessed so that

78
00:05:42,270 --> 00:05:46,170
the final comparison used specific features,

79
00:05:46,170 --> 00:05:47,910
whether letters or words:

80
00:05:47,910 --> 00:05:51,630
the same words occurring in the different samples.

81
00:05:51,630 --> 00:05:56,950
On that basis, we carried out the comparison between
the vectors.

82
00:05:57,310 --> 00:06:02,700
generated from these samples. In the first
experiment, we found that

83
00:06:02,700 --> 00:06:08,360
the clusters were unclear, and the samples had been
grouped

84
00:06:08,360 --> 00:06:12,980
incorrectly, because the text passages were very
short.

85
00:06:12,980 --> 00:06:16,720
We used about 500 words or fewer from each author.

86
00:06:16,720 --> 00:06:22,240
As you can see, the clusters were unclear. But in

87
00:06:22,240 --> 00:06:26,720
the second experiment, we increased the word count
until

88
00:06:26,990 --> 00:06:29,290
each sample had about 2,000 words.

89
00:06:29,290 --> 00:06:32,510
The clustering then looked different.

90
00:06:33,200 --> 00:06:38,340
As you can see, the clusters became very clear: the
texts grouped by

91
00:06:38,340 --> 00:06:43,780
author. This shows that it is possible to group
different

92
00:06:43,780 --> 00:06:49,300
writings by author using the words

93
00:06:49,300 --> 00:06:52,690
that occur most frequently.

94
00:06:52,690 --> 00:06:58,350
The second experiment is the most important to me,
and the subject of this presentation:

95
00:06:58,700 --> 00:07:01,720
comparing the Holy Quran with Arabic poetry.

96
00:07:01,720 --> 00:07:05,080
The Prophet, peace be upon him, was accused of being
a poet.

97
00:07:05,080 --> 00:07:08,660
The challenge was to produce something like this
Quran,

98
00:07:08,660 --> 00:07:10,300
and they failed to do so.

99
00:07:10,300 --> 00:07:14,740
They composed metrical verse, rhyming poetry, and so
on,

100
00:07:14,740 --> 00:07:17,760
but could not produce anything resembling the Quran.

101
00:07:18,020 --> 00:07:23,680
For this experiment, we collected poetry from the

102
00:07:23,680 --> 00:07:25,440
Pre-Islamic period to the Modern period.

103
00:07:27,600 --> 00:07:29,860
A very large collection, or corpus,

104
00:07:29,890 --> 00:07:34,750
so that we could compare these texts with the Holy
Quran.

105
00:07:35,760 --> 00:07:38,200
As you can see here,

106
00:07:38,200 --> 00:07:43,580
the corpus covers various periods, from Pre-Islamic,

107
00:07:43,580 --> 00:07:49,180
Early Islamic, Umayyad, Abbasid, Andalusian,
Ayyubid, Fatimid, and Mamluk,

108
00:07:49,180 --> 00:07:50,700
to Ottoman and Modern.

109
00:07:52,390 --> 00:07:58,300
We collected these groups of poems by different
writers.

110
00:07:58,720 --> 00:08:03,420
The corpus contained approximately

111
00:08:03,420 --> 00:08:05,940
ten million words,

112
00:08:05,940 --> 00:08:08,560
nearly six million lines of poetry,

113
00:08:08,560 --> 00:08:12,700
by around a thousand poets from different historical
periods.

114
00:08:13,480 --> 00:08:18,020
The first result, when we used the complete texts,

115
00:08:18,050 --> 00:08:23,070
was that, as you can see, there are two clusters.

116
00:08:23,230 --> 00:08:25,490
One contains the Holy Quran alone.

117
00:08:25,930 --> 00:08:31,230
The other, on the right, contains all the poetry
from the different periods.

118
00:08:31,230 --> 00:08:35,890
You can see that Umayyad poetry resembles
Pre-Islamic poetry,

119
00:08:35,890 --> 00:08:38,830
and these, in turn, resemble Early Islamic poetry.

120
00:08:38,830 --> 00:08:44,370
The other periods then group into different

121
00:08:44,370 --> 00:08:48,790
clusters. The Holy Quran is clearly distinct from
these texts.

122
00:08:49,020 --> 00:08:50,960
Then, when we took samples,

123
00:08:50,960 --> 00:08:54,680
we took samples from each period.

124
00:08:56,010 --> 00:09:01,810
The result was, again, that all

125
00:09:01,810 --> 00:09:07,530
ten samples from the Holy Quran clustered on their
own on the left.

126
00:09:07,530 --> 00:09:11,550
The remaining poetry clustered on the right.

127
00:09:12,160 --> 00:09:17,800
There is another very important point I'd like to
draw attention to:

128
00:09:17,800 --> 00:09:23,520
the ten Pre-Islamic poetry samples clustered close
to

129
00:09:23,520 --> 00:09:26,660
the Umayyad period, but still formed their own
group.

130
00:09:26,940 --> 00:09:30,680
This means they have distinctive characteristics.

131
00:09:30,910 --> 00:09:35,850
This proves that there really is Pre-Islamic poetry,

132
00:09:35,850 --> 00:09:38,390
contrary to what Professor Taha Hussein said.

133
00:09:38,390 --> 00:09:42,450
This issue should therefore be studied
scientifically.

134
00:09:42,450 --> 00:09:47,950
We should identify the features: the words, letters,
or

135
00:09:47,950 --> 00:09:50,130
the model used,

136
00:09:50,130 --> 00:09:53,310
on which this diagram was based,

137
00:09:53,310 --> 00:09:56,510
and which indicated that Pre-Islamic poetry has...

138
00:09:57,060 --> 00:10:00,740
distinctive features that set it apart from the
other periods.

139
00:10:01,020 --> 00:10:03,660
The important question, then, is:

140
00:10:03,660 --> 00:10:06,920
if we want to identify the features,

141
00:10:06,920 --> 00:10:11,060
the words used in these models and diagrams

142
00:10:11,060 --> 00:10:15,340
that distinguish the Quran from the other texts, and
study them,

143
00:10:15,340 --> 00:10:20,500
yes, we can do that, but using

144
00:10:20,500 --> 00:10:22,500
other methods.

145
00:10:22,890 --> 00:10:26,810
PCA is

146
00:10:26,930 --> 00:10:32,050
the subject of the next slide.

147
00:10:32,760 --> 00:10:38,320
With PCA, what we are trying to do is...

148
00:10:38,320 --> 00:10:43,860
Previously, we examined a text through the hundred

149
00:10:43,860 --> 00:10:49,200
words forming its vector, measuring its distance
from the other vectors.

150
00:10:49,480 --> 00:10:54,260
Now we want to look at these texts in just two
dimensions.

151
00:10:54,260 --> 00:10:59,400
That is why we use this technique.

152
00:10:59,400 --> 00:11:05,080
There are groups of words which

153
00:11:05,080 --> 00:11:09,240
occur more often in one group of texts and less
often in the other,

154
00:11:09,240 --> 00:11:14,480
according to the clusters. Here, for example, we
have the Quran on

155
00:11:14,640 --> 00:11:18,400
the left, surrounded by some of its distinguishing
words.

156
00:11:18,910 --> 00:11:23,370
On the right are the other poems and the words that
distinguish them.

157
00:11:23,370 --> 00:11:28,890
Some words are clear: for example, 'Allah' ('God'),
on the left beside the Quran.

158
00:11:28,890 --> 00:11:30,450
And 'al-dahr' ('time').

159
00:11:30,450 --> 00:11:35,750
These are clear. But other words, such as
demonstratives and so on,

160
00:11:35,750 --> 00:11:40,930
need input and analysis from language specialists to
understand

161
00:11:40,930 --> 00:11:46,250
why they occur more here than in the other texts,
and distinguish the Quran.

162
00:11:47,520 --> 00:11:51,490
If we move to two-word sequences,

163
00:11:51,490 --> 00:11:57,110
we again find different expressions in phrases or
pairs of words.

164
00:11:57,270 --> 00:12:01,810
The Holy Quran consistently differs in its
expressions,

165
00:12:01,810 --> 00:12:03,470
its words, and its vocabulary.

166
00:12:03,530 --> 00:12:09,090
If we leave that analysis and move to sequences of
letters,

167
00:12:09,090 --> 00:12:11,710
starting with single letters, for example,

168
00:12:11,710 --> 00:12:14,210
this is a result I think is very important.

169
00:12:14,210 --> 00:12:17,050
We find that a number of letters appear

170
00:12:17,270 --> 00:12:18,790
on the Quran's side

171
00:12:18,790 --> 00:12:20,830
more than in the poetry.

172
00:12:20,830 --> 00:12:23,690
The Quran used these letters more often.

173
00:12:23,690 --> 00:12:26,770
I don't know why. Perhaps it comes from vocabulary

174
00:12:26,770 --> 00:12:30,550
containing words that had been little used, or were
unique to it.

175
00:12:30,550 --> 00:12:33,410
They did not appear even in the later periods.

176
00:12:33,410 --> 00:12:36,430
It was distinguished by these letters.

177
00:12:36,430 --> 00:12:37,910
For example,

178
00:12:37,910 --> 00:12:40,190
here is a comparison,

179
00:12:40,190 --> 00:12:41,690
just to enlarge the picture,

180
00:12:41,690 --> 00:12:45,330
between the Quran, Umayyad, Early Islamic, and
Abbasid poetry. These are the same

181
00:12:45,720 --> 00:12:48,920
letters that distinguished it from all the other
periods:

182
00:12:48,920 --> 00:12:52,720
mim, kaf, nun, ha, and dhal.

183
00:12:52,720 --> 00:12:54,500
I don't know the explanation.

184
00:12:54,500 --> 00:12:56,000
We need language specialists

185
00:12:56,000 --> 00:12:57,580
to give us,

186
00:12:57,580 --> 00:12:59,000
as they say,

187
00:12:59,000 --> 00:13:01,650
an informed scholarly opinion.

188
00:13:01,650 --> 00:13:03,290
Likewise, if we use two letters,

189
00:13:03,290 --> 00:13:05,190
splitting the text into sequences

190
00:13:05,190 --> 00:13:09,790
of two letters, we again find the Quran distinct and
unique.

191
00:13:09,890 --> 00:13:12,310
If we move to three letters,

192
00:13:12,310 --> 00:13:14,410
we again find a different pattern.

193
00:13:14,890 --> 00:13:20,530
If we move to seven letters, there is something to
discuss here.

194
00:13:20,530 --> 00:13:26,310
I noticed that prepositions attach to nouns more
often in

195
00:13:26,310 --> 00:13:31,990
the Quran. But in Pre-Islamic poetry, as I'll show
on

196
00:13:31,990 --> 00:13:37,730
the next slide, they were very rare. What appeared

197
00:13:37,730 --> 00:13:42,990
more often, however, was prepositions with the
particle, or

198
00:13:42,990 --> 00:13:44,530
the definite article 'al-'.

199
00:13:44,850 --> 00:13:47,350
These appeared frequently in the Quran.

200
00:13:47,350 --> 00:13:52,630
I think the Quran then changed the way the language
was used, because other

201
00:13:52,630 --> 00:13:54,050
poetry began using them.

202
00:13:54,050 --> 00:13:59,830
Over time, it clearly began using prepositions
attached to 'al-'.

203
00:14:00,620 --> 00:14:06,370
There is a view that, previously, prepositions did
not

204
00:14:06,370 --> 00:14:09,310
attach directly to nouns in this way.

205
00:14:09,310 --> 00:14:12,650
The noun took the genitive, as in 'jamil al-wajh'
('beautiful of face').

206
00:14:12,650 --> 00:14:14,230
'Beautiful of face.'

207
00:14:14,230 --> 00:14:15,870
Or, as the poet said:

208
00:14:16,240 --> 00:14:18,360
'Whatever faults I may have,

209
00:14:18,360 --> 00:14:21,980
my dog is timid and my camel calf is lean.'

210
00:14:24,180 --> 00:14:26,740
The point is that the construction used to come like
this.

211
00:14:26,740 --> 00:14:31,260
This is one result that may support that view.

212
00:14:31,260 --> 00:14:36,680
But we also need scientific and technical analysis
from language

213
00:14:36,680 --> 00:14:41,360
specialists. This should be put to them for further

214
00:14:41,360 --> 00:14:46,020
discussion. In conclusion,

215
00:14:46,510 --> 00:14:49,850
the Holy Quran's style is completely different from
poetry.

216
00:14:49,850 --> 00:14:53,050
It cannot be poetry in any way.

217
00:14:53,050 --> 00:14:54,950
It differs in every characteristic,

218
00:14:54,950 --> 00:14:57,770
every feature, whether letters or words,

219
00:14:57,770 --> 00:15:03,530
or what we call parts or sequences of letters:

220
00:15:03,530 --> 00:15:05,750
two or three letters, and so on.

221
00:15:05,970 --> 00:15:11,050
The Quran's style is completely different from the
various poetic

222
00:15:11,050 --> 00:15:15,230
texts. And the Quran itself...

223
00:15:15,300 --> 00:15:20,940
The Quran itself says in Surah Ya-Sin: 'We have not
taught him poetry,

224
00:15:20,940 --> 00:15:23,760
nor is it fitting for him. It is only a reminder and
a clear Quran.'

225
00:15:23,800 --> 00:15:28,980
It cannot be poetry. In al-Haqqah: 'It is not a
poet's speech; little do you believe.'

226
00:15:29,000 --> 00:15:32,540
And in al-Shu'ara: 'Or do they say: a poet, for whom
we await the misfortunes of time?'

227
00:15:32,720 --> 00:15:35,860
Let them wait until the Hour comes.

228
00:15:36,060 --> 00:15:38,900
The Holy Quran will remain completely different,

229
00:15:39,000 --> 00:15:42,280
from all poetry, all prose, and everything else.

230
00:15:42,500 --> 00:15:44,800
It is a revelation from the Lord of all worlds.

231
00:15:44,860 --> 00:15:48,960
May God bless our Prophet Muhammad, his family and
companions, and grant them peace.

232
00:15:49,480 --> 00:15:51,880
Peace be upon you, and God's mercy and blessings.