A hybrid method of rule-based approach and statistical measures for recognizing narrators name in hadith

Hadith is one of the two main fundamental resources for Muslims which contains a collection of quotes that have been said by Prophet Mohammed. In order to validate Hadith, there are two main factors that can identify the strengthen of certain hadith which are the context (content of hadith itself) and the narrators (the persons who narrate this hadith). Identifying narrators' names plays an essential role in terms of validating specific hadith. In this research, we have used Named Entity Recognition which is a subfield of Natural Language Processing. Person name may yield a multi-word that indicates his first name such as Abdullah or in Arabic. This means that recognizing Arabic names requires to be treated as a multi-word expressions. Hence, this research addresses a method to recognize Arabic names from Hadith by a combination of rule-based and statistical methods. The process of this study consists of six phases which are dataset, transformation, pre-processing, Part-Of-Speech tagging, rule based method and statistical methods. The rule-based method is relying on a set of keywords which will trigger the start and end position of a narrator's name candidate. After the narator's name candidate is identified, it will be submitted to the statistical analysis to evaluate the possibility of the candidate as a narrator's name. The statistical measures that have been used are consisting of Log-likelihood Ratio (LLR), Point-wise Mutual Information (PMI), S-cost, R-cost and U-cost. The experimental results have reported an 86% of f-measure for the rule-based method, while LLR has outperformed the other statistical methods by obtaining an 85% of precision. In conclusion, the hybrid approach of rule based and statistical methods have provide a better result compared to relying only rule based method in recognizing narrator name is hadith.