Socio-economic database online

The SEDO project develops a platform combining fast access, user freedom, and coherence of the results for presenting complex socio-economic data. As an initial yield, this joint project involving two research actors of Luxembourg, the CEPS/INSTEAD for the content and the CRP-Gabriel Lippman for the technical realization, will deliver on the Net the results of longitudinal surveys about the life in Luxemburg in a generic tool comparable to those of the specialized community, characterized by flexibility and reutilization. Several search methods are available: a hierarchical browsing, a query engine with Boolean operators, and a top down navigation with minimal clicks for convenient and quick access to the main trends. Without the use of statistical tools or expertise in the domain the user can perform on the platform advanced statistical calculations. A new scheme for relational database has been designed to assure the portability of the application over different architectures. Last, the service manages the miscellaneous aspects of user identification, variable baskets, and the preparation of data for coherent analysis. 1 Networked access to socio-economic data The use of the Net for querying socio-economic databases has lately drawn the attention of both experts and the public at large, with a number of online services provided by statistical organizations of various countries. However, the volume of the data as well as the technical specialization of the domain curbs the customization and exploitation of the information [Br98]. Not only the networked publication of statistics gives the possibility of instantaneous update and correction, but it is also, more than the traditional paper communication, suited for building relationship with the data provider and the data consumer. The consumer can deliver his feedback and formalize his requirements, and the provider has means to control the usage of the data, so that the figures are better and more properly used in the practice. Other advantages are to ease the retrieval of large data sets, to allow their customization, and to enable end-users to execute on the server statistical functions for processing the data [Su97]. However, facilitating access to the information lead to specific problems such as selecting in the volume of the information, the pertinent data and operations made available to the public, and protecting the confidentiality of the data in whole. Infrastructure has to follow with the development of the functionalities and the evolution of the demand. When performing useful statistical analysis becomes easy, and leads to an increased demand, supporting exhaustive calculations simultaneously performed by an undefined number of clients forms a real technical challenge. 2 Related techniques The metadata, or data about data, are the specific information for describing the data considered as raw material for study or discussion. In the last decade, statistic frameworks have emerged from the development and normalization of metadata, so as to unify the collections of similar or correlated data [Me04]. As a result, various classifications, or views, of the raw data can be proposed according to the criteria of these metadata in such a way that static presentation of figure charts is no longer sufficient to fully render the information contained in the collection of statistics. Based on the metadata, two kinds of strategies for exploiting the data are proposed. The hierarchical approach proceeds by the exploration of an arborescent structuration of the data. The structure is represented in a Thesaurus: a conceptual lexicon organising abstract keywords in a graph of semantic dependences without loops. Searching for the information consists in walking the graph until the leaves where the effective data are stored. The enginesupported approach implements multi-criteria search, and proposes to the user a direct access to the data using his own keywords. Most services available on the Net implement both a hierarchical and a search-engine method. However, they often differ on the information area on which they are applied. For bibliographic search, engines are restricted to a collection of search criteria: period, localization, keywords, domains, author, etc... For access to statistical data, engine query leads to a list of survey variables with a brief description on the terms of the metadata, while the real results are kept in charts accessed by the Thesaurus.