AUTOMATIC QUESTION GENERATION FROM AMHARIC  SENTENCES USING A RULE-BASED APPROACH

MESRET, BEYENE ASCHALE

dc.contributor.author	MESRET, BEYENE ASCHALE
dc.date.accessioned	2024-03-05T08:44:20Z
dc.date.available	2024-03-05T08:44:20Z
dc.date.issued	2023-02
dc.identifier.uri	http://ir.bdu.edu.et/handle/123456789/15677
dc.description.abstract	Question generation is part of Natural Language Processing. Students today have numerous challenges when preparing for exams. Professors and Teachers spend a lot of time and effort to make exams. The Automatic Question Generation Model proposes a solution to save time, effort, and the student’s learning process, which helps in self-calibration for educational purposes. In this thesis work, we present an automatic question generation system for the Amharic language.AQG aims to develop a system that takes sentences of text and produces a good-quality question based on the text, such that the answer to the question can be worked out from the base sentences. The proposed system proceeds by transforming declarative sentences into interrogative sentences, based on preliminary named entity recognition of the base sentence. AQG generates questions automatically by using its model, which is generated using a rule-based approach.Question generation generates shallow questions that focus more on facts such as who/‹‹ማን››, what/‹‹ምን››, when/‹‹መቼ››, where/‹‹የት››, and why/‹‹ለምን››.We used Python programming language on Jupyter notebook Anaconda navigator which isa web-based interactive computing notebook environment withappropriate libraries. Our system is rule-based, runs on sentence-based parse information of a single-sentence input, and achieves high accuracy in terms of syntactic correctness and fluency. The AQG model uses pre-trained data from the open repository GitHub as a training dataset, which helps to get more accurate results. As a result, the question Generation systems rely on manual Evaluation. However, the proposed system was evaluated based on the quality of the output system and the linguistic well-formed type of criteria to evaluate the Result. Based on syntactic correctness 82.78% was generated and 88.78% was fluency generated questions. We also present an evaluation of the output system result, which shows that Recall was 82.35%,precisionwas 70.00% and F-measure was 75.67%. Which is good according to it generated automatically using rule based approaches. Keywords: Automatic Question Generation, Named entity Recognition	en_US
dc.language.iso	en_US	en_US
dc.subject	Computer Science	en_US
dc.title	AUTOMATIC QUESTION GENERATION FROM AMHARIC SENTENCES USING A RULE-BASED APPROACH	en_US
dc.type	Thesis	en_US