| Summary: | Fuzzy grammar has been introduced as an approach to represent and learn text fragments where the set of learned patterns are represented by combining similar segments to represent regularities and marking interchangeable segments. This paper is dedicated to present a procedural scheme towards learning text fragment in text categorization task facilitated by fuzzy grammars. A few issues are involved in developing fuzzy grammars which are (i) determination of the number of text classes to develop (ii) the selection of text fragments, F relevant to each text class (iii) determining frequent and important terms or keywords, V to develop the set of terminal, T and compound grammars, N. (iv) Conversion of text fragments into grammars, (v) Combination of grammars into a compact form. Comparison between fuzzy grammar and other location entity identifier such as LbjTagger, LingPipe, Newswire and Open Calais is observed where results have shown that this method outperforms other standard machine learning and statistical-based approach.
|
|---|