Loading...
CASOS DE ÉXITO
CDTI's knowledge graph powering intelligent management of technological knowledge
Starting point
The Centro para el Desarrollo Tecnológico Industrial (CDTI) faced a challenge common to many institutions with decades of history behind them: valuable information scattered across heterogeneous systems that did not communicate with one another. Structured files were held in Oracle databases whilst technical documentation lived in SharePoint, with no integrated way to access both repositories or to extract meaningful knowledge from combining them.
This fragmentation made it impossible to realise the true value of the accumulated technological heritage. Analysis required intensive manual work, searches were limited in scope, and strategic decisions lacked the support that decades of funded projects could have provided. The challenge was to transform this vast archive into a genuinely exploitable strategic asset.
The solution required going beyond simple systems integration. It was necessary to build a semantic layer that could understand the relationships between files, projects, organisations, technologies and people — turning isolated data into connected, actionable knowledge.
Solution overview
The knowledge graph as the system's core
The knowledge graph at the core of BICCO is not simply a unified database, but a semantic representation of the CDTI's technological universe. It is built using a Semantic Digital Model expressed in OWL/RDF — a formal language that allows machines to "understand" knowledge structures in a way similar to a human expert. This model can hybridise multiple ontologies to enrich the domain representation, capturing both the structured information from files and the results of automatic processing of technical reports using Natural Language Processing and Artificial Intelligence techniques.
Four fundamental properties define the power of the graph:
- Unification of heterogeneous information: Consolidates data from Oracle, SharePoint and external sources into a single semantic structure, eliminating the information silos that characterised the previous system.
- Queryability: Both people and machines can formulate complex queries against the accumulated knowledge, taking advantage of the explicit nature of semantic relationships.
- Knowledge expressiveness: Information is not limited to simple records, but is represented with the full richness of its contextual connections, enabling inferences and discoveries that would remain hidden in conventional structures.
- Extensibility: New entities, relationships or data sources can be incorporated without redesigning the entire system, simply by extending the existing semantic model.
Intelligent file search and exploration system
BICCO's semantic search engine goes beyond the limitations of traditional keyword-based search, incorporating predictive capabilities and a faceted search system that allows files to be filtered across multiple dimensions simultaneously.
The range of available facets reflects the richness of the semantic model: knowledge topics, specific technologies, economic sectors, timeframe, geographical location of development, file type and status, characteristics of participating organisations, funding programmes, aid purposes and modalities, areas of cooperation and budget ranges. This level of granularity means that a CDTI analyst can, for example, locate all biotechnology projects developed in Catalonia by SMEs over the past three years with budgets between €500,000 and €1,000,000 — in a matter of seconds.
Complementing these search capabilities, the system provides an enriched record for each file, incorporating contextual and recommendation systems grounded in named entity recognition, as well as dynamic graphical visualisations that aid understanding and analysis of the information.
Data enrichment system
BICCO's data enrichment system incorporates data into the knowledge graph and helps improve the user experience in relation to the file search and exploration system.
- Integration with external sources: The system connects with repositories from the Linked Open Data Cloud to contextualise files with information from DBpedia (the semantic version of Wikipedia), Wikidata, EPO Linked Data from the European Patent Office, DBLP for scientific publications in information technologies, and the Hércules project from the Research Management System of Spanish Universities. The latter is particularly valuable, as it has also been developed using GNOSS technology and is expected to publish data on researchers, publications, projects and patents.
- Automatic knowledge extraction: The system automatically reads the technical memory documents within each file, applies NLP techniques to recognise, disambiguate and link named entities, extracts technological descriptors and thematic topics, classifies files, and consolidates all of this information into the knowledge graph by linking it to existing entities.
Technology overview
To design the BICCO Knowledge Graph, a Semantic Digital Model has been developed that identifies the entities, attributes and relationships that make up the CDTI's knowledge universe. The Semantic Digital Model refers to that area of reality formed by a set of entities and their relationships, connected to the way people think about and engage with the world.
The Semantic Digital Model is a data representation model covering all knowledge objects within the CDTI's domain or area of application. It provides a map of the entities and sub-entities that make up all knowledge objects in that domain, along with the relationships between them. It takes into account both the content the model must represent and the knowledge objects that comprise it: the entities and attributes of the different knowledge objects, as well as the relationships between entities and between the different knowledge objects themselves. The model can hybridise multiple ontologies to provide a richer and more complete representation of the domain.
It is a system written in a formal language (OWL/RDF) that enables machines and systems to "understand" and correctly process the full set of entities described above — and in doing so, work alongside people within a common-sense framework in their processes of querying, information retrieval and knowledge discovery.
Impact
With BICCO integrated into the CDTI's working environment, the centre now has a more powerful and capable platform for carrying out intelligent searches across unstructured data — flexibly, simply, and with full customisation — without compromising the confidentiality of the information managed, in strict compliance with the CDTI's data security and confidentiality policies.
- Decision-making tool: BICCO has become a cornerstone of the centre's decision-making processes. Management can now ground strategic direction in rigorous analysis of accumulated knowledge, identifying emerging technology trends, detecting funding gaps in specific areas, and evaluating the historical impact of different aid programmes.
- Intelligent and flexible search: The ability to combine multiple filtering criteria has transformed the day-to-day work of analysts. What previously required days of manual queries across different systems is now resolved in seconds, with greater precision.
- Unified, accessible knowledge: The integration of heterogeneous information into a single intelligent access point has eliminated the information silos that characterised the previous environment, enabling cross-cutting analyses that relate files, organisations, technologies and outcomes in ways that were previously impractical.
- Process automation: The automatic extraction of information from technical reports has freed up specialist human resources, which can now be directed towards higher-value tasks rather than the manual classification and tagging of documents.
- Contextual enrichment: Connections with external sources significantly broaden the context of each file, making it possible to relate funded projects to the current state of scientific knowledge, relevant patents or complementary university research lines.
- Guaranteed scalability: The knowledge graph and formal semantic model architecture allows the system to be extended with new entity types or data sources without redesigning the entire structure, ensuring that BICCO can evolve alongside the centre's needs.
- Security and confidentiality: All of these capabilities have been implemented in strict compliance with the CDTI's data security and confidentiality policies. The system incorporates granular role- and permission-based access controls that ensure each user can only access the information relevant to their role.
- Improved operational efficiency: Project evaluation and monitoring processes have been significantly streamlined. Assessors can quickly consult the technological history of applicant organisations, identify synergies with previous projects or detect potential duplication — all through a single, intuitive interface.
- Technological evolution: BICCO establishes an open platform ready to incorporate new tools and future approaches. Its modular architecture allows emerging generative AI techniques, predictive analytics or advanced visualisation to be integrated without compromising existing operational components.