Semi Rule-Based Poem Meter Classification of Recited Verses. Masters thesis, King Fahd University of Petroleum and Minerals.

PDF SemiRuleBased_Meter_classification_signed.pdf - Accepted Version Restricted to Repository staff only until 8 April 2027. Download (8MB)

Arabic Abstract

يُعَد ُّالشِّعْر ُأَحَد َأَقْدَم ِأَشْكَال ِالفُنُون ِالمُسْتَخْدَمَة ِحَوْل َالعَالَمِ، وَقَد ِاسْتُخْدِم َلِلتَّعْبِير ِعَن ِالمَشَاعِر ِوَنَقْل ِالأَحَاسِيس ِوَتَسْجِيل ِالأَحْدَاث ِالتَّارِيخِيَّةِ. وَيُشَكِّل ِعِلْم ُالعَرُوض ِالإِطَار َالَّذِي يَضْبِط ُبِنْيَة َالقَصِيدَة ِالعَرَبِيَّة ِوَيُوَجِّه ُإِلَى كِتَابَة ِالشِّعْر ِالسَّلِيمِ. لَقَد ْبُذِلَت ْجُهُود ٌعَدِيدَة ٌفِي مَجَال ِتَصْنِيف ِأَوْزَان ِالشِّعْرِ، وَتَرَكَّزَت ْأَحْدَثُهَا عَلَى اسْتِخْدَام ِالشَّبَكَات ِالعَصَبِيَّة ِالمُتَكَرِّرَة ِأَو ْنَمَاذِج ِالمُحَوِّلَاتِ. وَتُرَكِّز ُهَذِه ِالدِّرَاسَة ُعَلَى الأَسَالِيب ِالتَّقْلِيدِيَّة ِفِي تَصْنِيف ِالأَوْزَان ِوَكَيْف َيُمْكِن ُتَطْبِيقُهَا بِاسْتِخْدَام ِالتِّقْنِيَّات ِالحَدِيثَةِ. وَمِن ْأَبْرَز ِالتَّحْسِينَات ِالَّتِي قَدَّمَهَا هَذَا البَحْث ُالتَّحَوُّل ُمِن ْنِظَام ِالتَّرْمِيز ِالتَّقْلِيدِي ِّإِلَى التَّمْثِيل ِالثُّنَائِيِّ، ْمِمَّا أَلْغَى الحَاجَة َإِلَى التَّقْطِيع ِالعَرُوضِي ِّالتَّقْلِيدِي. وَلَقَد ْتَم َّتَقْيِيم ُعِدَّة ِأَنْظِمَة ٍلِلتَّشْكِيل ِالتَّلْقَائِي ِّعَلَى نُصُوص ِالشِّعْر ِالعَرَبِيِّ. وَلِلتَّعْوِيض ِعَن ِالأَخْطَاء المُحْتَمَلَة ِفِي التَّشْكِيلِ، اسْتُخْدِم َنَهْج ٌمُبْتَكَر ٌيَعْتَمِد ُعَلَى المُحَاذَاة ِالزَّوْجِيَّة ِلِحِسَاب ِالتَّشَابُه ِفِي التَّرْمِيزِ. وَقَد ْحَقَّق َالمَنْهَج ُالمُقْتَرَح ُدِقَّة ًإِجْمَالِيَّة ًبَلَغَت 77.69% عَلَى مَجْمُوعَة ِبَيَانَات ٍنَصِّيَّة ٍو 78.63َ% عَلَى مَجْمُوعَة ِبَيَانَات ٍصَوْتِيَّةٍ. وَبِالإِضَافَة ِإِلَى ذَلِكَ، أَظْهَر َتَحْلِيل ُالنَّتَائِج ِأَن َّبَعْض َالأَخْطَاء ِفِي التَّصْنِيف ِنَاتِجَة ٌعَن ْأَخْطَاء ٍفِي وَسْم ِالبَيَانَاتْ، مِمَّا يُشِير ُإِلَى الحَاجَة ِإِلَى التَّحَقُّق ِالجَيِّد ِمِن ْصِحَّة ِمَكَانِز ِالبَيَانَات ِالعَامَّةِ.

English Abstract

Poetry is a traditional form of art used throughout the world. It has been utilized to share emotions, convey feelings, and record historical events. Arabic prosody is the science that governs the structure of poems and guides into writing sound poetry. Multiple efforts were put into poetry meter classification. Most recent efforts focus on using Recurrent Neural Networks or Transformer models. This research focuses on the traditional methods of meter classification and how they can be implemented using modern technologies. A major upgrade from the traditional symbolic encoding scheme is the change to a binary sequence which eliminated the need for the traditional sequence segmentation. Multiple automatic diacritization systems were evaluated on Arabic poetry text. To account for errors in the diacritization, a novel approach of using pairwise alignment to calculate sequence similarity was used. The proposed approach achieved a total accuracy of 77.69\% on a text-based dataset and 78.63\% on a speech-based one. Additionally, analyzing the results showed that some of the misclassifications were caused by errors in the labeling which indicates that proper validation of public datasets is needed.

Item Type:	Thesis (Masters)
Subjects:	Computer
Department:	College of Computing and Mathematics > Information and Computer Science
Thesis Advisor:	Wasfi Al-khatib,
Thesis Committee Members:	Moataz Ahmed, Jawad Abuihlail,
Depositing User:	MURTADHA ALABDULSALAM (g202007520)
Date Deposited:	09 Apr 2026 08:48
Last Modified:	09 Apr 2026 08:48
URI:	http://eprints.kfupm.edu.sa/id/eprint/144074