In the field of multi-application personalization, several techniqueshave been proposed to support user modeling foruser data management across different applications. Manyof them are based on data reconciliation techniques oftenimplying the concepts of static ontologies and generic userdata models. None of them have sufficiently investigatedtwo main issues related to user modeling: (1) profile definitionin order to allow every application to build their ownview of users while promoting the sharing of these profilesand (2) profile evolution over time in order to avoid datainconsistency and the subsequent loss of income for web-siteusers and companies.In this paper, we conduct work and propose separated solutionsfor every issue. We propose a flexible user modelingsystem, not imposing any fixed user model whom differentapplications should conform to, but based on the concept ofmapping among applications (and mapping functions amongtheir user attributes). We focus in particular on the managementof user profile data propagation, as a way to reduce theamount of inconsistent user profile information over severalapplications.A second goal of this paper is to illustrate, in this context,the benefit obtained by the integration of a Semantic Layerthat can help application designers to automatically identifypotential user attribute mappings between applications.This paper so illustrates a work-in-progress work wheretwo complementary approaches are integrated to improvea main goal: managing multi-application user profiles in asemi-automatic manner.