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：2013
卷号：54
期号：2
出版社：Journal of Theoretical and Applied
摘要：Most of the crimes committed today are reported on the Internet through news, blogs and social networking sites. These sources have provided a huge amount of crime data, presenting a need for a means to extract useful information. In this research, the evaluation of Direct and Indirect extraction of nationality from crime news, along with the additional references to identify the nationalities of suspects, victims and witnesses is presented. Named entity recognition using gazetteers and rule-based extraction are used, in addition to co-reference resolution to link references. The proposed approach was evaluated and compared to manual extraction system. The results indicate that the proposed approach is able to extract most information related to nationality from crimes news and identify the additional information references. Its performance proved good, with 55% precision, 96% recall, and 70% f-measure evaluation metrics.
