Abstract:
This study incorporates automatic question answering (QA) system, which is one of the
most important sub-fields of Natural Language Processing (NLP). Here, the system
comprises two phases. In the first phage, the system generates four automatic options of
the answer for the input question based on the given topic. On the other phase, the system
provides multiple choice questions where an examinee can answer the questions within a
time constraint. Based on the examinees answers he/she will get a score. It may be
mentioned that the question may be multiple choice and descriptive, but this report only
deals with factoid or multiple-choice questions. Both parts of the research were
implemented for Bangla Language which is one of the low resource languages. Because of
the scarcity of resources, the research based on Bangla Language does not reach up to the
mark. Nowadays the availability of internet resource for the Bangla Language attract
researchers for doing research at different dimensions though the accuracy depends on
whether the corpus is domain oriented or generalized. For domain-oriented corpus, the
accuracy of the application is very good but the application based on the generalized corpus
provides accuracy which is not satisfactory in some cases. It is observed from the system
investigation that the first and the second phases provide the low and high accuracy
respectively based on the domain-oriented data. The originality of the project is phase-I
which generates four options based on the input question for Bangla Language by using
the n-gram language model.
