Morphology-based entity and relational entity information extraction framework for Arabic -

Abstract

Natural Language Processing is concerned with automating the understanding of natural language. Morphological analysis is key to Arabic natural language processing due to the morphological richness of Arabic. Researchers proposed and evaluated knowledge-based and empirical techniques to extract entities and relations from text. Knowledge-based techniques require advanced linguistic and programming expertise. Empirical techniques require large training and reference corpora to learn and evaluate computational models respectively. In this work, we present a morphology-based entity and relational entity information extraction framework for Arabic text. The framework provides a user-friendly interface where the user defines tag types and associates them with regular expressions defined over Boolean formulae. Boolean formulae are terms, negations of terms, and disjunctions of terms where terms are matches to Arabic morphological features. The framework introduces a semantic feature that relates words based on synonymity. The framework allows the user to associate an action with each regular sub-expression and to define semantic relations. The framework uses an in-house Arabic morphological analyzer to compute morphological matches, computes regular expression matches, and then builds the relations across matches. We evaluated our work with several case studies and compared with existing application-specific techniques.