SYSTEM AND METHOD FOR LARGE-SCALE ARABIC LEXICAL SEMANTIC ANALYSIS

摘要

System and method for extracting word senses sequences (01 ) corresponding to sequences of input Arabic words (11 ). These senses belong to a global compact basis set of predefined semantic fields. The system can also produce the semantic relations (02) between the words of two input Arabic sequences of words (11, 12). It relies on a lexical semantic relational database that relates the lexical Arabic compounds to semantic fields both in the forward and backward directions. To achieve high coverage of the highly derivative and inflective Arabic language, this database is not a vocabulary but a morpheme based one. Semantic fields are associated with lexical compounds with the aid of a large-scale morphological analyzer. Semantic relations between words are determined by first reducing the words to lexical compounds, mapping lexical compounds to semantic fields, and relating the latter to each other. This approach reduces complexity considerably.