Using SVM Algorithm to Improve the Extraction of Arabic Composite Names: A Case Study in the Economic Domain

Abstract

The performance of natural languages has been developed by the unique processing applications (Named Entity Recognition) this is why the task of recognition has been of special significance. Therefore, this paper focuses on information extraction from Arabic unstructured text. Named entity recognition is a critical subtask of information extraction; it is the process by which a system can automatically detect and categorize Named Entities (NE). The proposed work in this research is looking for how to extract composite Arabic names in the economic domain by using machine learning approach specifically an SVM algorithm. To implement and test the proposed work the GATE tool was used. Also, for evaluation purposes and measure the efficiency and accuracy of the experiments, several measures were used including Precision, Recall and F- measure rates. Finally, a comparison was made between the results of the proposed system with the results obtained from previous research that depends on the extraction of composite Arabic names using rule-based approach. The application of SVM to extract Arabic composite named was successful. Results show that the proposed model based on machine learning in terms of Precision for extracting composite named is higher than the results obtained from the rule-based method. On the contrary, the obtained results of Recall of rule-based approach were higher than the results obtained from the machine learning SVM model.