Isemantec
The iSemantec project was developed in the context of the issues of knowledge capitalization, management and
reuse among large industrial corporations.
In these corporations, data and documents related to their products and coming from different actors are already integrated by systems called "Product lifecycle management" (PLM).
However, this integration is based on the formalization of the main processes of the corporation. And this is often a detriment to the specificity of the different professions. Hence, those professions have established their own databases and documents beyond the capitalization.
Furthermore, in a context of markets instability, it seems necessary, in order to meet client needs, to adapt the information system to continuous reconfigurations of processes and partners networks.
A "Product lifecycle management" system can be seen as a specialized document system, in which documents are organized according with the structure of the product parts they deal with (cf. Fig. 1).
A major problem for the flexibility required by professions specificity and processes reconfigurations is that the types of parts and relations are "hard-wired" in the PLM relational database model, and cannot be changed without information technologists' intervention. To solve this, our partner, Cadesis, designed a "metamodel" for PLM, and specializations for different projects or partners (cf. Fig. 2).
Cadesis developped a software called "pdm2rdf" in order to extract the data from a major PLM software and convert it to RDF triples, while defining as a RDF schema the links between the particular models to the metamodel. Both were stored in a RDF framework (Jena), allowing them to be queried through a SPARQL service (Joseki). Our laboratory was in charge with allowing users themselves to add emerging data. Because there is still no standard way to update RDF data, we designed an integration architecture (see Fig. 3) which uses:
- rdf2hypertopic, a web service adapter we developped to convert hypertopic requests into SPARQL requests on-the-fly,
- a new version of Agorae which we developped so that it can integrate data coming from different hypertopic servers.
Through this architecture, Agorae can query and display both read-only stabilized data from PLM projects and read-write emerging data stored in Argos (see Figure 4, 5).
Owing to an initial delay in the release of data from our partner ABB (931,338 triples for 37,987 items), we first used the platform on the data of another Cadesis client (824 triples for 83 items). The differences between those two data sets helped us test the genericity and the scale-up ability of the platform.
Some queries appeared to be nearly independent of the number of items (e.g. the detail of the item, the items types), whereas other queries took much longer (e.g. the list of attributes used in the base, the list of values used for an attribute). This scale-up problem resides in fact in the RDF layer (whether SPARQL is implemented by Joseki or RDF API). In order to get usable response times, we had to simplify some queries (e.g. the list of attributes defined in the base), and had to implement a cache.
A realistic use scenario was established by our laboratory and a quality manager from ABB. To illustrate capitalization and flexibility, several qualitative viewpoints (see Fig. 6) were created for three types of actors who felt under-represented in the current PLM: a project manager, a quality manager and someone from the purchase service (see Fig. 7 and 8).
To introduce a bit of intersubjectivity in this industrial use case, we provided asynchronous awareness. Every web page from Agorae is now associated with an Atom feed (generated by Argos) related to the operations done on the corresponding item, viewpoint or topic. Hence the user can visualize the history of one of them (see Fig. 9), and can choose to be notified on future updates.
Chao Zhou, Scaling up the Socio-semantic Web: Application protocol and user interfaces, PhD thesis, Université de technologie de Troyes, 2009.
Aurélien Bénel, Ontologies du Web : Histoire refoulée et perspectives paradoxales, Intellectica 2014/1(61), 123–141. ARCO, 2014.