The automotive industry is undergoing a profound digital transformation. Electric vehicles, advanced driver assistance systems (ADAS), connected vehicle platforms, autonomous driving technologies, predictive maintenance solutions, and AI-powered customer support systems all depend on one critical asset: data.

Yet data alone is not enough.

Modern automotive ecosystems generate enormous volumes of information every day. Vehicle manufacturers, suppliers, software developers, engineers, regulators, translators, and AI systems continuously create and exchange technical knowledge. The challenge is no longer collecting information. The challenge is ensuring that everyone interprets it in exactly the same way.

Although all three are referring to the same concept, they do not necessarily understand it in the same way. As automotive organizations become increasingly dependent on AI, this lack of shared understanding creates significant risks.

Terminology inconsistencies can affect machine translation quality, reduce the reliability of AI-generated responses, complicate data sharing between organizations, and limit interoperability between software systems.

This challenge was the starting point for EcoDrive TermSpace, a European Data Space developed between March 2025 and May 2026 by Pangeanic, Universitat Jaume I (UJI), and ValgrAI.

The project's objective was not simply to create another terminology database. It sought to establish a structured knowledge environment where automotive concepts, terminology, and linguistic assets could be represented, shared, and reused consistently across organizations and languages.

Why Automotive AI Needs More Than Data

Most Artificial Intelligence systems learn from large collections of text. Large Language Models (LLMs) analyze billions of words to identify patterns and predict likely responses. While this approach has produced remarkable results, it also has limitations.

Language models learn correlations between words. They do not inherently understand the meaning of the concepts those words represent.

This distinction becomes particularly important in the automotive industry, where accuracy is critical. An incorrect interpretation of a maintenance procedure, a mistranslated safety instruction, or an inaccurate AI-generated recommendation can have operational, regulatory, legal, or safety implications.

To address this challenge, AI systems require more than language data. They require structured knowledge.

What is an Ontology ?

An ontology is a formal representation of knowledge within a specific domain. Rather than storing isolated words and definitions, it captures:

• Concepts that exist within a field
• Categories to which those concepts belong
• Relationships connecting concepts
• Rules governing interactions between concepts

In effect, an ontology creates a structured model of a domain that can be interpreted consistently by both humans and machines.

From Automotive Terminology to Automotive Knowledge

One of the central outcomes of EcoDrive TermSpace was the development of an ontological framework for the automotive domain using the ONTODIC methodology developed by the Universitat Jaume I research team.

The ONTODIC model transforms terminology resources into interconnected knowledge structures. Instead of treating terms as isolated database entries, ONTODIC represents them as instances of broader concepts connected through semantic relationships.

For example, the concept of an electric vehicle battery can be linked to battery management systems, charging infrastructure, energy storage technologies, vehicle diagnostics, and safety procedures.

These connections allow AI systems to understand not only the concept itself but also its place within a wider network of automotive knowledge.

How Ontologies Improve Machine Translation and Generative AI

One of the most practical applications of the EcoDrive TermSpace ontology is its ability to support both Machine Translation (MT) and Generative AI systems.

Traditional machine translation systems often rely on statistical probabilities or neural language patterns to determine how a term should be translated. While highly effective in many situations, these systems may struggle with specialized terminology, emerging technologies, or context-dependent concepts.

A common limitation of Large Language Models is hallucination: generating responses that appear plausible but are factually incorrect. This occurs because language models generate text based on probability rather than verified knowledge.

By connecting AI systems to a structured knowledge source such as EcoDrive TermSpace, organizations can provide a trusted reference layer containing validated automotive concepts and relationships.

To demonstrate this capability, the project developed an AI-powered virtual assistant connected directly to the underlying Data Space. Instead of relying solely on model memory, the assistant accesses structured automotive knowledge represented through the ontology.

Building the Knowledge Infrastructure for Enterprise AI

For much of the past decade, discussions about AI have focused primarily on models. Organizations have invested heavily in larger datasets, larger models, and greater computational capabilities.

Increasingly, however, another layer is emerging as a strategic differentiator: knowledge infrastructure.

Ontologies, semantic frameworks, knowledge graphs, and interoperable data spaces provide the structure that enables AI systems to move beyond pattern recognition and operate with greater contextual understanding.

As AI becomes embedded in automotive engineering, manufacturing, mobility services, customer support, and autonomous systems, the ability to combine language intelligence with structured knowledge may become one of the most important foundations of reliable enterprise AI.