The "logicist Corpuses" project has three complementary and interdependent aims, serving knowledge cumulativity in the field of human sciences. The first is to develop methods and tools for constituting so-called "logicist" corpuses, i.e. corpuses consisting of documents structured in data and interpretation rules, the latter being understood as inference operations performed to generate conclusions or interpretative hypotheses. These corpuses will have a double function: as a decision-making aid to help researchers with their scientific interpretation, and a documentary function to manage and share all the data used to support the interpretations proposed. Thus they will directly contribute to the process of knowledge accumulation as well as to a research dynamic. The second is to constitute, based on this model, archaeology of techniques corpuses, a field of excellence in Europe. The third aim is to develop an automatic annotation tool based on an ontology enabling queries to be effected on logicist corpuses both on the interpretation rules and data.

To satisfy these aims it is planned first of all a) to transform into logicist documents a significant number of scientific texts concerning the archaeology of techniques, b) collect and index data connected to these scientific constructs, c) translate these documents in English or French (depending on the original language), d) assess the rules produced in terms of transferability in order to produce corpuses consisting of rules with a "universal" or "local" character. In parallel, as far as publishing tools are concerned, it is planned to develop 1) annotation tools, and 2) consultation interfaces. Rules and data annotation will be done semi-automatically thanks to the use of automatic language processing and semantic annotation tools based on an ontology. The rule and data reading interface will comprise a) a natural language interrogation device, b) an editing device for the answers giving a reading by levels, in the sense that it will possible to read the different rules relating to the question asked, and then their premises, thereby checking the solidity of the available rules and data by consulting the original publications.

The scientific community expects: a rapid reading of the rules used by the researchers to obtain or support a result, easily understandable scientific reasoning, and, in return, better sharing of knowledge within the discipline, exhaustive access to the databases on which the scientific constructions of a field are based, the auto-archiving of research data and a solution for the perpetuation of the indexing of the data. The technologist community is especially sensitive to these expectations, in so far as the present publishing process does not allow experimental data to be generally shared even though these data are indispensable to the dynamic of their research.

Furthermore, the constitution of logicist corpuses in the field of technology should serve as a model for the type of corpus which could be developed in the field of human sciences, and could in this way play a direct role in an effective accumulation of knowledge.