Service Communautaire d'Information sur la Recherche et le Développement - CORDIS

Contextualization engine

The Context Engine is responsible for contextualising textual logical resources and SVG historical maps. It includes the transformation engine for rendering contextualised textual resources and supporting the contextualisation of visualisations that can utilise XML-based contextualisation data, such as the SVG Historial maps used in the VICODI project.

The context engine includes both a server and client. Training data submitted to the server are used to create Context Index Vectors, which are correlation vectors, which describe relationships between entities (concepts or instances) known to the system. In the present system, only terms URIs (concept or instance) derived from any ontology are sent by the client to the server. The server records the ontology entities received during training and also their top level concept e.g. location, category, person, event, artefact, abstract notion, social organisation, etc -

The server not only supports training, but also vector similarity functinalities: term-term, term-document, document-document. High-level functionality includes:
- Provides automated context estimate for given resource term frequency vector.
- Create training resource given expert term vector and frequency vector. The expert ranks each term in resource according to the importance to the resource.
- Provides context ranking (IR) for given list of result context vectors and one comparison vector.

The client also supports LATCH reseasoning utilising the context of the resource (dynamically derived or stored) and the ontology. LATCH context means that we predict the most important locations, categories and time intervals (narrow for historical map queries, etc. and a wider interval for user reference).

Informations connexes

Reported by

See on map