Proverb Variation: Experiments on Automatic Detection in Brazilian Portuguese Texts

摘要

This paper describes a methodology for automatically identifying proverbs and their variants in running texts. This methodology is based on existing compilations of proverbs, by exploring the regular syntactic structures that most proverbs present and intersecting syntactic structure with the lexical units of the proverbs. From the syntactic regularities we divided the data into 13 different classes. Finite-state automata is used to represent the regular patterns found in the classes. The results showed a precision rate of 74.68% tested in Brazilian Portuguese journalistic corpus.