Military

Abstract—Parsing systems able to analyze natural language text robustly would be of great value in computer applications such as Information Retrieval, Knowledge Extraction and automatic translation. In real world applications, the parser must deal efficiently with difficult input that cannot be parsed correctly according to the standard grammar rules in the system, whether it is an extra-grammatical sentence, ill-formed or unexpected input. Most systems use algorithmic approaches to robustness where parsing programs are extended to include heuristics to handle defect cases. In this work we adopt another solution based on a tolerant grammar-based approach for developing a robust parser for unrestricted Arabic sentences. It consists of introducing robust rules in the grammar itself and relaxing constraints if necessary. The implemented parser lends itself to applications where large scale text processing is involved and robust, reliable, and relatively accurate syntactic analysis is necessary. The system is based on a wide coverage linguistic knowledge from the DIINAR-MBC European project1. The parser has been evaluated against real-world sentences and the results were very encouraging. The parser provides 95 % coverage