Using NLP for semi automated class diagram generation from Arabic text

العناوين الأخرى

استخدام أدوات معالجة اللغة الطبيعية للأتمتة الشبه آلية لتصميم مخطط الفئة من نص مكتوب باللغة العربية

مقدم أطروحة جامعية

al-Khatib, Lubnah Ghazi

مشرف أطروحة جامعية

Tartir, Samir

الجامعة

جامعة فيلادلفيا

الكلية

كلية تكنولوجيا المعلومات

القسم الأكاديمي

قسم علم الحاسوب

دولة الجامعة

الأردن

الدرجة العلمية

ماجستير

تاريخ الدرجة العلمية

2017

الملخص الإنجليزي

Arabic language is one of the most used languages in the world.

Software requirements of an information system are written in natural language; Arabic or English.

Arabic is the formal language that is used to write software requirements by the most of ministries and governmental institutions in Arab countries.

In software engineering, manual analysis and design of the requirements cause variations of the obtained model due to the complexity of natural language.

Using natural language processing tools in analysing and transforming requirements into models enhances knowledge extraction performance.

Many approaches addressed this domain, but very little research has been developed to obtain models from requirements written in Arabic, especially class model.

Also, they never use semantic approaches that utilize the domain ontologies, which produce a gap between knowledge engineering and software engineering.

In this thesis, we propose an approach that solve the previous problems and automate the process of constructing a class model from Arabic requirements.

The core aim of our approach is to handle Arabic user requirements transformation into class diagrams.

We propose a set of identification rules to identify class diagram elements, then we combine the proposed set of identification rules, ANLP tools and domain ontologies to build a unique approach for generating UML class diagram from Arabic user requirements.

Our approach is the first approach that utilizes domain ontology and Arabic natural language processing tools to generate class diagrams form the Arabic requirements.

Finally, we validate our approach by performing an experiment, which involves a group of human evaluators.

The results showed that our approach produces very good results as compared to the evaluators depending on the performance evaluation criteria; precision and recall.

The results indicated that our proposed technique is a good candidate for extracting class diagrams from Arabic user requirements text as compared to existing approaches.

التخصصات الرئيسية

تكنولوجيا المعلومات وعلم الحاسوب

عدد الصفحات

51

قائمة المحتويات

• APA
• MLA
• AMA

لغة النص

الإنجليزية

نوع البيانات

رسائل جامعية

رقم السجل

BIM-955502