Exploiting Wikipedia to support exploratory Arabic search on the web

العناوين الأخرى

استغلال الويكيبيديا لتدعيم البحث الاستكشافي العربي في الويب

مقدم أطروحة جامعية

Abid, Ahmad Muhammad Abd al-Aziz

مشرف أطروحة جامعية

al-Agha, Iyad Muhammad

أعضاء اللجنة

Abu-Shaban, Yusuf Nabil
al-Halis, Ala Mustafa

الجامعة

الجامعة الإسلامية

الكلية

كلية تكنولوجيا المعلومات

القسم الأكاديمي

تكنولوجيا المعلومات

دولة الجامعة

فلسطين (قطاع غزة)

الدرجة العلمية

ماجستير

تاريخ الدرجة العلمية

2016

الملخص الإنجليزي

Due to the huge amount of data published on the Web, the Web search process has become more difficult, and it is sometimes hard to get the expected results, especially in case of explanatory search when users are unfamiliar with the search domain.

Many efforts have been proposed to support exploratory search on the Web by using different knowledge sources such as DBpedia and Linked Open Data (LOD).

However, these knowledge sources have limited support for the Arabic content, and thus they can be hardly used with queries expressed in Arabic.

In this research, we propose a fully automated approach that is run on query time to support search results for Arabic language by exploiting Wikipedia link structure.

It aims to use the Arabic version of Wikipedia to extract complementary knowledge that is relevant to the search query submitted by the user.

We propose ArabXplore, a system that extracts key entities from search snippets and Wikipedia pages and ranks them based on a new ranking algorithm that is based on the traditional PageRank algorithm.

Finally, a graph is built to visually represent highly ranked topics and their relations to the end user.

Our proposed system was assessed over a dataset of 100 Arabic search queries covering different domains, and results were assessed and rated by a human expert.

The underlying ranking algorithm was also compared with the conventional PageRank.

Results showed that our ranking algorithms outperformed the PageRank algorithm.

Our ranking algorithm achieved 87.7 nDCG and 68.2 MAP while the conventional PageRank achieved 84.5 nDCG and 50.3 MAP.

The source code, test dataset, and complete experimental results are available online on: https://github.com/aabed91/ArabXplore

التخصصات الرئيسية

تكنولوجيا المعلومات وعلم الحاسوب

عدد الصفحات

77

قائمة المحتويات

• APA
• MLA
• AMA

لغة النص

الإنجليزية

نوع البيانات

رسائل جامعية

رقم السجل

BIM-727250