東華大學圖書館 |

Measuring Semantic Similarity for Arabic Sentences Using Machine Learning = = قياس مستوى التشابه بين الجمل العربية بإستخدام التعلم الآلي .‪‪‪

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Measuring Semantic Similarity for Arabic Sentences Using Machine Learning =/
其他題名:	قياس مستوى التشابه بين الجمل العربية بإستخدام التعلم الآلي .‪‪‪
作者:	Alian Abdel Rahim, Marwah Mustafa.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2021,
面頁冊數:	228 p.
附註:	Source: Dissertations Abstracts International, Volume: 83-12, Section: B.
Contained By:	Dissertations Abstracts International83-12B.
標題:	Artificial intelligence. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28863637
ISBN:	9798819377659

Measuring Semantic Similarity for Arabic Sentences Using Machine Learning = = قياس مستوى التشابه بين الجمل العربية بإستخدام التعلم الآلي .‪‪‪
Alian Abdel Rahim, Marwah Mustafa.

Measuring Semantic Similarity for Arabic Sentences Using Machine Learning =قياس مستوى التشابه بين الجمل العربية بإستخدام التعلم الآلي .‪‪‪ - Ann Arbor : ProQuest Dissertations & Theses, 2021 - 228 p.

Source: Dissertations Abstracts International, Volume: 83-12, Section: B.

Thesis (Ph.D.)--Princess Sumaya University for Technology (Jordan), 2021.

This item must not be sold to any third party vendors.

Semantic similarity represents an important task in many natural language processing applications. In addition, semantic similarity is a metric that shows the relatedness among words, sentences, or documents according to the likeness of their meaning. Semantic similarity has several applications, such as plagiarism detection, text entailment, text summarization, information retrieval, machine translation, and paraphrasing identification. The contribution of this dissertation is the improvement of semantic similarity measurement between sentences using word semantic analysis. This method prepared a good representation for words while considering their context, part of speech, and senses. A good representation for words in the distributional space is provided by word embeddings models such as Word2Vec, but senses and word order are not considered. To overcome the problem of polysemy, we investigated multi-sense representation by measuring the semantic similarity between sentences and use chunks to overcome the word order limitation of Word2Vec. The Arabic language has specific features, such as morphological features, word order, and words with multi-senses that make handling the Arabic text a challenging task. This study is based on word embedding to measure semantic similarity between Arabic sentences while taking into account the impact of several factors, such as part of speech, chunks, and sense representation. To prove the results of the proposed approach, we applied the method on paraphrasing identification. Then, we use human ratings with Pearson correlation and automatic evaluation with precision and recall metrics for evaluation. In addition, we developed an Arabic dataset for paraphrased sentence pairs to be used in the evaluation process. The sentence similarity is measured by including syntactic, semantic, and syntactic-semantic features and then applying linear regression to learn the coefficients of these features and predict the similarity between tested sentences. Experiments are conducted using two datasets, namely, our generated Arabic paraphrasing benchmark and MSRvid (SemEval 2017). The proposed approach achieved a correlation of 0.743 on the SemEval 2017 dataset. However, when applied to the Arabic paraphrasing benchmark, the approach achieved a correlation of 0.354 due to the complexity of this dataset.

ISBN: 9798819377659Subjects--Topical Terms:

516317
Artificial intelligence.
Subjects--Index Terms:

Chunks

Measuring Semantic Similarity for Arabic Sentences Using Machine Learning = = قياس مستوى التشابه بين الجمل العربية بإستخدام التعلم الآلي .‪‪‪
LDR:07227nmm a2200421 4500 001 2352167
005 20221118093830.5
008 241004s2021 ||||||||||||||||| ||eng d
020 $a 9798819377659
035 $a (MiAaPQ)AAI28863637
035 $a AAI28863637
040 $a MiAaPQ $c MiAaPQ
100 1 $a Alian Abdel Rahim, Marwah Mustafa. $3 3691790
245 1 0 $a Measuring Semantic Similarity for Arabic Sentences Using Machine Learning = $b قياس مستوى التشابه بين الجمل العربية بإستخدام التعلم الآلي .‪‪‪
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2021
300 $a 228 p.
500 $a Source: Dissertations Abstracts International, Volume: 83-12, Section: B.
500 $a Advisor: Awajan, Arafat.
502 $a Thesis (Ph.D.)--Princess Sumaya University for Technology (Jordan), 2021.
506 $a This item must not be sold to any third party vendors.
520 $a Semantic similarity represents an important task in many natural language processing applications. In addition, semantic similarity is a metric that shows the relatedness among words, sentences, or documents according to the likeness of their meaning. Semantic similarity has several applications, such as plagiarism detection, text entailment, text summarization, information retrieval, machine translation, and paraphrasing identification. The contribution of this dissertation is the improvement of semantic similarity measurement between sentences using word semantic analysis. This method prepared a good representation for words while considering their context, part of speech, and senses. A good representation for words in the distributional space is provided by word embeddings models such as Word2Vec, but senses and word order are not considered. To overcome the problem of polysemy, we investigated multi-sense representation by measuring the semantic similarity between sentences and use chunks to overcome the word order limitation of Word2Vec. The Arabic language has specific features, such as morphological features, word order, and words with multi-senses that make handling the Arabic text a challenging task. This study is based on word embedding to measure semantic similarity between Arabic sentences while taking into account the impact of several factors, such as part of speech, chunks, and sense representation. To prove the results of the proposed approach, we applied the method on paraphrasing identification. Then, we use human ratings with Pearson correlation and automatic evaluation with precision and recall metrics for evaluation. In addition, we developed an Arabic dataset for paraphrased sentence pairs to be used in the evaluation process. The sentence similarity is measured by including syntactic, semantic, and syntactic-semantic features and then applying linear regression to learn the coefficients of these features and predict the similarity between tested sentences. Experiments are conducted using two datasets, namely, our generated Arabic paraphrasing benchmark and MSRvid (SemEval 2017). The proposed approach achieved a correlation of 0.743 on the SemEval 2017 dataset. However, when applied to the Arabic paraphrasing benchmark, the approach achieved a correlation of 0.354 due to the complexity of this dataset.
520 $a يمثل التشابه الدلالي جزء مهما في العديد من تطبيقات معالجة اللغات الطبيعية. إنه مقياس يوضح العلاقة بين الكلمات أو الجمل أو المستندات وفقًا لتشابهها من حيث المعنى. . التشابه الدلالي له العديد من التطبيقات مثل اكتشاف الانتحال، والاستلزام في النص، وتلخيص النص، واسترجاع المعلومات، والترجمة الآلية، وتحديد الترادف وغيرها. في هذا البحث ، تتمثل مساهمتنا في تحسين قياس التشابه الدلالي بين الجمل باستخدام التحليل الدلالي للكلمات. يتطلب ذلك إعداد تمثيل جيد للكلمات مع مراعاة سياقها واعرابها والمشترك اللفظي لبعض الكلمات. تم توفير تمثيل جيد للكلمات من خلال تضمين الكلمات (Word2Vec) ولكنه لا يأخذ في الاعتبار المشترك اللفظي وترتيب الكلمات. للتغلب على مشكلة المشترك اللفظي ، فإننا نبحث في تمثيل المشترك اللفظي كمتجه في قياس التشابه الدلالي بين الجمل بينما يتم استخدام تحليل الجملة الى أجزاء للتغلب على مشكلة ترتيب الكلمات في Word2Vec . تتميز اللغة العربية بسمات محددة مثل السمات الصرفية وترتيب الكلمات والكلمات متعددة المعنى (المشترك اللفظي) التي تجعل التعامل مع النص العربي مهمة صعبة. نهجنا المقترح هو نهج قائم على تضمين الكلمات لقياس التشابه الدلالي بين الجمل العربية مع مراعاة تأثير عدد من العوامل مثل اعراب الكلمات، وتحليل أجزاء الجملة، وتمثيل المشترك اللفظي. لإثبات نتائج النهج المقترح، تم تطبيق طريقتنا في تحديد الترادف وتم تقييمها باستخدام التقييمات البشرية حسب قانون ارتباط بيرسون والتقييم التلقائي بدقة ومقاييس الاسترجاع. بالإضافة إلى ذلك، قمنا بإنشاء قاعدة بيانات عربية لأزواج الجمل المعاد صياغتها لاستخدامها في عملية التقييم. يقاس تشابه الجملة من خلال استخدام السمات النحوية والدلالية والنحوية- الدلالية ثم تطبيق linear regression لمعرفة معاملات هذه الميزات والتنبؤ بالتشابه بين الجمل المختبرة. يتم إجراء التجارب باستخدام اثنتين من datasets؛ مجموعة الجمل المعاد صياغتها والتي تم إنشاؤها من قبلنا و MSRvid -SemEval 2017. حقق النهج المقترح ارتباطًا قدره 0.743 في MSRvid datasetبينما حقق ارتباطًا قدره 0.354 عند تطبيقه على dataset للجمل المعاد صياغتها بسبب التعقيد الناتج عن استخدام قواعد التحويل في هذه الجمل.‪‪‪
590 $a School code: 2116.
650 4 $a Artificial intelligence. $3 516317
650 4 $a Information technology. $3 532993
650 4 $a Language. $3 643551
650 4 $a Linguistics. $3 524476
650 4 $a Middle Eastern studies. $3 3168421
653 $a Chunks
653 $a Dependency parsing
653 $a Part of speech
653 $a Semantic similarity
653 $a Sense representation
653 $a Word embedding
690 $a 0800
690 $a 0489
690 $a 0555
690 $a 0679
690 $a 0290
710 2 $a Princess Sumaya University for Technology (Jordan). $b Department of Computer Science. $3 3682737
773 0 $t Dissertations Abstracts International $g 83-12B.
790 $a 2116
791 $a Ph.D.
792 $a 2021
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28863637