AUTOMATIC METHOD FOR EXTRACTING THE RELEVANT PHRASES FROM TEXTS.

摘要

The present invention refers to an automatic method for extracting the relevant phrases from texts without reading them upon analyzing the entropy characteristics of the information, also applying the Pareto statistics. The entropy analysis of the information is useful for separating the hazard (disorder) from the order (relevance), discriminating the words that do not contribute to the meaning from texts, but which are part of the grammatical structure of the file. The Pareto statistic is subsequently applied for obtaining the extreme behaviour of the arrangement of the relevant words in texts, this latter with the purpose of establishing categories of relevance which help the user of the information with the classification of large amounts of files, thereby showing in an organized table the most relevant words or phrases, which provide meaning to the analyzed text. This method is applied to any language and does not require processing the analyzed texts. The inventive method avoids the use of experts in the thematic areas and the definition of lexicon or dictionaries, thus reducing costs and increasing the speed in analysing texts.