期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2018
卷号：96
期号：23
出版社：Journal of Theoretical and Applied
摘要：Several data sources like social networks and blogs are providing increasing amounts of unstructured data in natural language. This data contains useful information that must be identified automatically, quickly and with high precision. Therefore, different Information Extraction (IE) approaches were proposed like Rule-based and Statistical ones. The majority of rule learning IE systems present a recurring problem caused by the generation of a set of irrelevant and unnecessary rules, which affects the quality of the extraction results. Hence, we propose in this paper a novel and generic approach to increase the performance of these extractors in order to avoid missing important information or providing erroneous one. It consists in enhancing the rule generalization through using a domain ontology designed to make the systems able to generate only the most likely useful rules. To prove its efficiency, our solution is applied to (LP)2 system and empirically tested on a corpus for seminar announcements. According to our results, the system�s enhanced version reaches a high accuracy with respect to (LP)2 and other extractors, which means that the information it extracted is of a better quality.
关键词：Information Extraction; Rule-Based Learning; Natural Language Processing; Ontology
