Arithmetic Word Problem Solver Using Frame Identification

Abstract

AutomaticWord problem solving has always posed agreat challenge for the NLP community. Usually a word problemis a narrative comprising of a few sentences and a questionis asked about a quantity referred in the sentences. Solvingword problem involves reasoning across sentences, identificationof operations, their order, relevant quantities and discardingirrelevant quantities. In this paper, we present a novel approachfor automatic arithmetic word problem solving. Our approachstarts with frame identification. Each frame can either beclassified as a state or an action frame. The frame identification isdependent on the verb in a sentence. Every frame is unique and isidentified by its slots. The slots are filled using dependency parsedoutput of a sentence. The slots are entity holder, entity, quantityof the entity, recipient, additional information like place, time.The slots and frames helps to identify the type of question askedand the entity referred. Action frames act on state frame(s) whichcauses a change in quantities of the state frames. The frames arethen used to build a graph where any change in quantities canbe propagated to the neighboring nodes. Most of the currentsolvers can only answer questions related to the quantity, whileour system can answer different kinds of questions like ‘who’,‘what’ other than the quantity related questions ‘how many’.There are three major contributions of this paper:1) Frame Annotated Corpus (with a frame annotation tool).2) Frame Identification Module.3) A new easily understandable Framework for wordproblem solving.