Loading...
What is a knowledge graph?
Knowledge graphs are the artificial mind of information systems.
The challenge of the digital age: from data swamps to intelligent data
The challenge of the digital age: from data swamps to intelligent data //
In today's digital age, organisations face an unprecedented challenge: managing massive quantities of data that are growing exponentially. And this figure continues to rise with the expansion of digitalisation across all sectors.
The main challenge lies not only in the storage and processing of these vast volumes of information, but in the ability to establish meaningful connections between apparently disparate data in order to understand and correlate them. Information, though abundant, remains fragmented and disconnected, limiting its potential to drive innovation and informed decision-making.
"From the very beginning, information technologies have turned their back on meaning"
This tendency persists in current natural language processing technologies and large language models (LLMs). These models are self-referential; they generate humanised text based on statistical patterns, but lack any true understanding of meaning because they have no symbolic capacity. In other words, they are unable to recognise the truth or falsity of the propositions they produce.
A truly artificial intelligence must have a commitment to truth, which implies a relationship with something external to itself; it must incorporate a symbolic relationship that connects with reality. Current generative technologies, however, relate only to themselves and therefore lack any reference to the external world. Strictly speaking, they represent nothing.
Against this backdrop, knowledge graphs emerge naturally as a tool that provides an effective solution to this challenge, enabling organisational information to be structured, connected, and mined for value in a systematic and scalable way.
These relationships, which act as bridges between facts, define how the different elements within the graph are connected and enable the inference of new knowledge from existing connections. In this way, knowledge graphs link and unify information meaningfully, making it queryable by both humans and machines.
For people, it is straightforward to distinguish between entities such as Luisa and Juana, or between Logroño and Madrid. From this recognition, we attribute to each entity what belongs to it: that Logroño is in La Rioja, or that Juana was born in Madrid. This process constitutes, for us, the act of knowing: understanding relationships and contexts. The condition for building artificial intelligence is that our systems recognise those same entities and establish relationships between them.
A graph is, in essence, a mathematical structure composed of nodes and edges, where nodes represent entities and edges reflect the relationships between them. Knowledge graphs, however, go beyond this basic structure: they must be modelled using an ontology which, through the establishment of clear rules and constraints, enables context to be interpreted and dispersed data to be transformed into consolidated, structured facts — turning them into knowledge.
Knowledge graphs contain and distinguish, thanks to their logical foundation, the real-world entities of a domain or field of knowledge: they are capable of reasoning, drawing inferences, and learning, making them fundamental tools for advanced artificial intelligence applications.
Such applications include, for example:
- High-precision recommendation systems, which understand the relationships between users, products, services, and their preferences.
- Information retrieval and semantic search systems, which improve the relevance and ranking of results by interpreting context and the connections between search terms.
- Natural language processing, where the ability to structure and contextualise data enables deeper and more precise understanding.
Knowledge graphs possess numerous capabilities that radically transform our relationship with data and lay the foundations for artificial intelligence with genuine contextual understanding.
True knowledge occurs at the messy intersection of ideas
A Knowledge Graph brings together and connects the multiple human perspectives on a topic and provides personalised responses.
| They unify distributed information | They make information queryable by both machines and people, and increase its expressiveness and extensibility |
| They understand and interpret | They enable domain-specific data to be integrated, achieving a richer and more contextualised representation of information. Their structure facilitates the interconnection of heterogeneous and distributed data, improving accessibility and interoperability, without ambiguity or loss of information |
| They query and find | They enable the implementation of natural language querying systems, with more intuitive and cognitively ergonomic queries, and their combination with faceted search, offering multiple linked and contextual ways of accessing and exploring data |
| They reason, discover, and learn | They enable the creation of an ecosystem of linked and interconnected data that is easily accessible. This approach drives a smarter web, improving search processes, knowledge discovery, and the personalisation of content according to users' specific needs. They can serve as context to help computers understand and manipulate data, and provide a knowledge base that makes NLP techniques and large language models more accurate, useful, and intelligent. |
| They enhance the user experience | They promote a framework of cognitive ergonomics in the relationship between humans and systems, contributing to a space of digital wellbeing. By connecting natural language querying systems with faceted search, access to information is democratised, making it more accessible to a wider audience. Users can discover and explore any topic in a deeper and more intuitive way, and enjoy a semantically aware web. |
Capabilities of knowledge graphs operated and enriched with AI
| Automatic Metadata Tagging and Categorisation | Automatic interpretation of the entities modelled in the knowledge graph, endowing them with contextual meaning and organising them automatically. This enables every element of the knowledge graph to be comprehensible to both machines and people, facilitating its discovery and its linking with other related entities. |
| Digital Ecosystem | Integration of heterogeneous and distributed data held in different silos — representing all the agents within the knowledge domain — into a unified digital ecosystem within a semantically interpreted knowledge graph, through which information flows meaningfully, creating a coherent and enriched experience for people. |
| Single Point of Query | An intelligent metasearch engine that enables users and managers to access any entity and find exactly what they are looking for, regardless of where the source information resides, providing contextualised and personalised responses. |
| Contextualisation of resources in space and time | Intelligent visualisation of the graph's entities on interactive maps and timelines, enabling people to discover entities by geographical or temporal proximity. |
| Context Generation | The ability to create personalised narratives that connect different entities according to the individual's profile, interests, and preferences. |
| High-Precision Recommendations | A system that suggests recommendations based on explicit preferences, but also on a meaningful and in-depth understanding of user behaviour, implicit interests, and the context of their activity. |
| Knowledge Discovery | The ability to infer new connections and opportunities for exploration through automatic reasoning, surfacing new information from the implicit content of the graph and identifying non-obvious activity patterns. |
| Dynamic Visualisation for Management | Real-time dashboards that enable both managers and users to visualise and understand activity within the domain modelled in the graph from multiple perspectives. |
| Natural language querying system | A natural language querying capability able to hold conversations on any aspect of the knowledge domain modelled in the graph, proposing — on the basis of hosted content enhanced with Generative AI capabilities — auditable responses to complex questions and providing contextualised recommendations based on the comprehensive knowledge supplied by the graph. |
| Proactive Information Management | A communication system that proactively delivers information of potential interest to individuals, enabling anticipatory and forward-looking management. |
| Distributed Authoring with Dynamic Semantic Publishing | Enables all actors within the system to contribute their own content, which is automatically integrated into the digital ecosystem, keeping it consistently up to date and coherent. |
| Auditable Artificial Intelligence | Alignment with the principles of auditable and explainable AI for compliance with regulations such as the European AI Act and the ENIA, which promote transparent, ethical, and people-centred artificial intelligence. In systems where errors cannot be tolerated, the output of the AI must provide a truth that is auditable, traceable, and reproducible by third parties. |