Inconsistency detection in Islamic advisory opinions using multilevel text categorization

Abstract

Inconsistency detection is a large research area that has many applications. In the scope of Islamic content mining, this topic is of a particular interest because of the continuously increasing content and the need of people to find out more about its authenticity. Inconsistency detection is usually performed using linguistic analysis as well as the application of logic rules. We propose here a new method for inconsistency detection based on multilevel text categorization. For each categorization level, discriminative keywords are extracted using the hyper rectangular decomposition method which outputs the keywords in a hierarchical rank of importance. Then, those keywords are fed into the random forest classifier which automatically detects the category of each advisory opinion. Inconsistency detection is performed using an algorithm that detects inconsistent paths of advisory opinions. This study has been validated on a set of Islamic advisory opinions related to vows. The results are very interesting and show that our method is very promising in the field.