Definition Question Answering System for Afan Oromo Language

Dejene, Hundessa (2015) Definition Question Answering System for Afan Oromo Language. Masters thesis, Addis Ababa University.

[img] PDF (Definition Question Answering System for Afan Oromo Language)
Dejene, Hundessa.pdf - Accepted Version
Restricted to Repository staff only

Download (1MB) | Request a copy

Abstract

The amounts of electronic documents on the web are increasing explosively from time to time. Users were relying on IR system to explore their information of interest.IR cannot satisfy today’s users’ information need because it only returns a ranked list of documents rather than a precise and exact answers for users’ needs. Question answering system is a solution to provide exact answers for users’ information need with the help of information extraction techniques. The demand of users for retrieving precise information from the electronic Afan Oromo documents is increasing. An attempt had been done to develop factoid question answering for this language. In this study we have designed Afan Oromo Definition Question Answering system. The system architecture had been specified and its major components are question analysis, document processing and answer extraction. Pattern based approach had been used to extract target word from users’ natural language question and definition from the corpus. We have found that the question analysis and answer extraction components are the major language dependent component of the system. And also we have noticed that the usage of stemming and synonyms list have a great effect on the performance of the system. Using the surface pattern approach, the definition extraction component correctly extracts 93.3% of the definiendum from users’ question. Only 6.7% have been wrongly identified. The standard performance measures metrics recall, precision and F-measure was used for evaluation. We have performed the experiment with and without applying the stemming and synonyms. Accordingly without the usage of stemming and synonyms the system performance was precision 85%, recall 58.72% and F- measure 69.45%.And with the application of stemming and synonym the performance of the system was recall 78%, precision 86.4% and F-measure 81.9%. Usage of surface pattern approach for answer extraction from Afan Oromo documents had shown an encouraging result. Preparing patterns in detail for compound words more improve the performance of the system.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Question answering, Afan Oromo Question Answering, Definition question.
Subjects: H Social Sciences > H Social Sciences (General)
P Language and Literature > PL Languages and literatures of Eastern Asia, Africa, Oceania
Z Bibliography. Library Science. Information Resources > Z665 Library Science. Information Science
Z Bibliography. Library Science. Information Resources > ZA Information resources
Divisions: Africana
Depositing User: Selom Ghislain
Date Deposited: 01 Oct 2018 12:00
Last Modified: 01 Oct 2018 12:00
URI: http://thesisbank.jhia.ac.ke/id/eprint/5828

Actions (login required)

View Item View Item