Word/phrase based answer type classification for Bengali question answering system

A Comprehensive Comparison of Machine Learning Based Methods Used in Bengali Question Classification

“…We use different types of classifiers for QA type classification. We separate our methodology into two sections similar to [14] -one is training section and another is validation section shown in Figure 2.…”

“…The set of question categories is known as question taxonomy [14]. We have used two layer taxonomy which was proposed by Xin Li, Dan Roth [17].…”

“…This two layer taxonomy is made up of two classes which are Coarse Class and Finer Class. There are six coarse classes such as Numeric, Location, Entity, Description, Human and Abbreviation and fifty finer classes such as city, state, mountain, distance, count, definition, group, expression, substance, creative, vehicle etc as shown in the Table I [14]. A coarse-grained description of a system denotes large components while a fine-grained description denotes smaller sub-components of which the larger ones are composed of.…”

“…In this system, we choose bi-gram for extracting features. 3) Stop Words: We use two setups (as done in [14]) for our system. In the first setup, we eliminate the stop words from the text using another dataset containing only stop words.…”

“…The goal of QC is to assign classes accurately to the questions based on expected answer. In modern system, there are two types of questions [14]. One is Factoid question which is about providing concise facts and another one is Complex question that has a presupposition which is complex.…”