1. UKM Learning and Research Repository
2. Master Thesis / Tesis Master
3. Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Title:	Arabic question classification using support vector machine and pattern matching
Authors:	Ali Mutalleb Hasan (P72241)
Supervisor:	Lailatul-Qadri Zakaria, Dr.
Keywords:	Question classification Vector machine Pattern based techniques Arabic Hadith Universiti Kebangsaan Malaysia -- Dissertations
Issue Date:	26-Sep-2016
Description:	Question classification plays a crucial role in the question answering system. The goal of question classification is to accurately assign one or more labels to question based on expected answer type. Nonetheless, classifying user's question is a very challenging task due to the flexibility of Natural Language where a question can be written in many different forms and information within the sentence may not be enough to effectively to classify the question. Limited researches have focused on question classification for Arabic question answering. Previous researchers use handcrafted rules and keyword matching for question classification. However, the methods cannot be easily adapted to a new domain. This research propose a combination model based on support vector machine and pattern based techniques for Arabic question classification Hadith domain question answering in this research also studied the affect of a feature set on the performance of support vector machine for question classification . Five patterns were introduced to analyze and classify three types of questions which are Who, Where and What. The dataset set used in this research consist of 200 question about Hadith from Sahih Al Bukhari. The experimental result scored F-measure at 88.39%, 87.66% and 87.93% respectively for Who"
Pages:	85
Publisher:	UKM, Bangi
URI:	https://ptsldigital.ukm.my/jspui/handle/123456789/476447
Appears in Collections:	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.