A hybrid method for extracting relations between Arabic named entities

Abstract

Relation extraction is a very useful task for several natural language processing applications, such as automatic summarization and question answering. In this paper, we present our hybrid approach to extracting relations between Arabic named entities. Given that Arabic is a rich morphological language, we build a linguistic and learning model to predict the positions of words that express a semantic relation within a clause. The main idea is to employ linguistic modules to ameliorate the results that are obtained from a machine learning-based method. Our method achieves encouraging performance. The empirical results indicate that the hybrid approach outperformed both the rule-based system (by 12%) and the machine learning-based approaches (by 9%) in terms of the F-score, to achieve 75.2% when applied to the same standard testing dataset, ANERCorp.

Keywords

• Hybrid method
• Relation extraction
• Named entity
• Machine learning
• Genetic algorithm
• Rule-based method