Compression textuelle sur la base de règles issues d'un corpus de sms (Textual Compression Based on Rules Arising from a Corpus of Text Messages) [in French]

Textual Compression Based on Rules Arising from a Corpus of Text Messages The present research seeks to reduce the size of text messages on the basis of compression techniques observed mostly in a corpus of sms. This paper explains the methodology followed to establish compression rules. It then presents the 33 considered rules, and illustrates the four suggested levels of compression with two practical examples, automatically generated by a first prototype. This research’s main purpose is not to produce "sms-language", but consists in designing a textual compression process able to generate short and understandable texts from any textual source in French. The term of "essentialization" is proposed to describe this approach of textual reduction. MOTS-CLEFS : résumé automatique, compression de texte, sms, lisibilité, essentialisation.