Small Scale Linguistic Models to Improve Email Filtering and Representation

O. Nouali (Algeria), P. Blache, A. Regnier (France), and A. Mokhtari-Aissani (Algeria)


Information filtering, neural network, machine learning, reduced linguistic models, user model and text architecture.


Email software usually propose mail filtering devices. This filtering is a lexical property based classification. The presence or absence of keywords that the user enters to the software according to fields of the email dispatch the emails toward an appropriate folder, which contains all the emails that share this or those properties. This article proposes an improvement of these mail filtering: The proposed method expands the basic model with a set of linguistic properties, thus achieving higher prediction performance. These properties are based on reduced linguistic models. They can be automated and are considered as a set of clues that deal with the mail structure and content.

Important Links:

Go Back