A Method For Extracting Three Kinds Of Implicit Quantity Relations From Algebra Problems

Recently, researchers pay more and more attention to machine solver. This paper aims at extracting implicit relations from algebra problems. The main work is made up of 3 parts.Firstly, classifying the implicit relations. The implicit relations are made up of common sense relations, math format relations and units' relations. Considering the differences,different methods are used in different kind of relations.Secondly,constructing an implicit relation database. It is of great practical significance in the machine solver. It can increase the speed of the machine solver and make maintenance of the database simple. This method can also be used in another subject in future.Lastly, using the classifying technology and keyword matching method to extract implicit relations. Classifying technology and keyword matching method are used to judge whether the problems need this kind of implicit knowledge or not. Then we only need to build a mapping between characters of relations and the words in the problems.This paper implemented the extracting of implicit relations. Combining the extracting results with explicit relations, the machine solver can get over almost all of the algebra problems. The model is tested in the problems gathered from textbooks. The precision rate is over 90 percent. It declares that the model can be applied in the machine solver and it can also be applied in another subject in the future.