Abstract-In the area of data management for information system and especially at the level of e-learning platforms, the Big Data phenomenon makes the data difficult to deal with standard database or information management tools. Indeed, for educational purposes and especially in a distance training or online research, the learner that uses the e-learning platform is left with a heterogeneous set of data such as files of all kinds, curves, course materials, quizzes, etc. This requires a specialized fusion system to combine the variety of data and improve the performance, robustness, flexibility, consistency and scalability, so that they can provide the best result to the learner The user of the e-learning platform. In this context, it is proposed to develop a tool called "Big-Learn" based on a technique to integrate the mixing of structured and unstructured data in one data layer, and, in order to facilitate access more optimal search relevance with adequate and consistent results according to the expectations of the learner. The methodology adopted will consist initially in a quantitative and qualitative study of the variety of data and their typology, followed by a detailed analysis of the structure and harmonization of the data to finally find a fictional model for their treatment. This conceptual work will be crowned with a working prototype as a tool achieved with UML and Java technology.