NAFIS: A Gold Standard Corpus for Arabic Stemmers Evaluation

Arabic stemming as an important pre-processing task in Arabic natural language processing services and applications experience two serious deficiencies: "unique stemming solution" and "stemmers' performance inconsistency". These defects are mainly caused by the absence of a Gold Standard Corpus. Defined as a collection of texts stored in an electronic format, selected to be representative of a particular language, collection or genre, manually annotated and enriched with additional linguistic information, such corpus is used in stemmers benchmarking works. This paper provides a sight on NAFIS (Normalized Arabic Fragments for Inestimable Stemming), an Arabic stemming gold standard corpus. We describe NAFIS building methodology and we use it as an evaluation corpus in a benchmarking exercise.