Lisbon's linguists built the search algorithms powering AI today

In a narrow corridor on the fourth floor of a building in the Alfama district, a researcher in 2025 was debugging an ontology-matching algorithm. The code integrated lexical signals, contextual features, and language-model-based embeddings into what its documentation described as a "unified, interpretable architecture." The repository was called Exact-OM. It lived on GitHub under the account liseda-lab. And its institutional home was a university that, in 2026, had spent nearly a century studying what language is and how it works.

This is the story of the Lisbon School not a formal academy with a charter and enrollment, but a loose lineage of researchers, centers, and ideas that grew up around the question of what words mean and how machines might come to know it too. It is a story told in Portuguese, in computational notations, in repositories, and in the particular way a small country's scholars learned to think big.

The philology room that became a language laboratory

The Center of Linguistics of the University of Lisbon known by its Portuguese initials, CLUL traces its founding to 1932. In its first years, it operated as a center of philological studies, the traditional European practice of studying texts and languages in their historical contexts. The work was literary, archival, and rooted in the belief that understanding Portuguese required understanding how Portuguese had been spoken, written, and transformed across centuries.

By 1976, CLUL had evolved. It had developed into a research center in linguistics that expanded its activities and scientific scope over time, and was integrated as a centre of the University of Lisbon in 2004. The shift was significant: from philology as text interpretation to linguistics as a broader science of language knowledge, acquisition, and use. This reorientation set the stage for what would come later the moment when a philology room meets a computer lab.

In 2008, CLUL became a research center of the School of Arts and Humanities of the University of Lisbon, positioned within the area of Language Sciences. Its mission had grown to encompass everything from the mental representation of grammar to oral, audiovisual, and written language processing. The center's scope extended beyond linguistics itself, drawing in disciplines of psychology, cognitive science, medical sciences, speech therapy, computational engineering, history, anthropology, philosophy, and literature.

The scope CLUL claims for itself is striking in its ambition: it studies Portuguese in its multiple varieties diachronic, social, regional, and national and Portuguese-related Creoles, Portuguese as native, non-native, and heritage language, in monolingual, bilingual, multilingual, and contact settings, across typical and atypical populations. The center frames Portuguese as having unique characteristics in the Romance language space that offer "an unrivalled opportunity for cross-linguistic and typological studies to uncover the specific and universal properties of language."

"We are an interdisciplinary research center within the Portuguese science system committed to advancing the understanding of language knowledge, acquisition and use."

That interdisciplinary commitment the willingness to sit at the intersection of linguistics, computer science, and cognitive psychology would become the defining characteristic of the Lisbon approach to semantic search.

Three rooms and a specialized library

CLUL houses three laboratories for experimental linguistics: the Psycholinguistics Lab, the Phonetics and Phonology Lab, and the Lisbon Baby Lab, which studies early language acquisition. An additional speech lab facility handles high-quality acoustic recording. The center also owns a specialized library of over 30,000 volumes.

The numbers tell part of the story: 86 integrated members with PhDs, 20 integrated members without PhDs, 53 collaborators, 8 current financed projects, and 5 research groups. But the more important fact is what these rooms and people are building toward: the scientific study of language at a scale and precision that only becomes possible when a center has accumulated decades of expertise, equipment, and institutional memory.

The connection to modern search and discovery technology is not immediately obvious until you consider what semantic search actually requires. To match a query to a document, to understand that "Lisbon" in one context refers to the same city as "Lisboa" in another, to recognize that "semantic search" and "meaning-based retrieval" point to the same concept these tasks require precisely the kind of deep linguistic knowledge that centers like CLUL have spent decades developing.

Portuguese, in CLUL's framing, offers "unique characteristics in the Romance language space." This matters because Romance languages present particular challenges for computational processing: rich morphology, flexible word order, extensive inflection, and idiomatic expressions that resist simple rule-based parsing. Working on Portuguese meant developing techniques that could handle linguistic complexity and those techniques, once refined, traveled well to other languages.

The Center of Linguistics of the University of Lisbon sees service to the community as a central part of its research activities, and develops initiatives to broadly disseminate research outcomes and promote their application to societal demands, especially in the domains of Language and Speech Technologies, Health Care and Well-being, Education,, and Social Inclusion. Language and Speech Technologies sits at the top of that list a direct bridge to the computational work that would later define semantic search.

The computational turn: LISEGA and the semantic data laboratory

If CLUL represents the linguistic foundation, the Lisbon Semantic Data Lab represents what happened when that foundation met modern computer science. The lab known by its GitHub handle liseda-lab is a research group part of LASIGE at Faculdade de Ciências, Universidade de Lisboa. The contact email points to a researcher at ciencias.ulisboa.pt, the science faculty of the University of Lisbon.

LASIGE is a research unit in informatics and engineering, the kind of place where computer scientists build systems and test them against real problems. The Lisbon Semantic Data Lab takes the questions that linguists like those at CLUL have spent decades asking what does meaning look like? how do words connect to each other? and translates them into computational terms.

The lab's repository listing reads like a catalog of contemporary semantic search challenges. KGE_Predictions_GD is a project for "Predicting Gene-Disease Associations" using knowledge graph embeddings. Kgsim-benchmark handles benchmarks for knowledge graph similarity. VOWLMap works with visual notations for ontologies. ML4ReferenceAlignment tackles machine learning for reference matching. REx focuses on "Rewarding Explainability in Drug Repurposing with Knowledge Graphs."

The projects span domains from bioinformatics to drug discovery, but they share a common technical substrate: knowledge graphs, ontology matching, semantic similarity, and the kind of structural reasoning that makes semantic search possible. These are not toy problems. They represent the cutting edge of how computers are being taught to understand relationships between concepts and to make those relationships searchable.

One repository, Exact-OM, deserves particular attention. It is described as a "Hybrid ontology matching framework that integrates lexical, contextual, and lm-based signals within a unified, interpretable architecture." Ontology matching is one of the foundational technical challenges in semantic search: when different databases or knowledge systems use different terms for the same concept, how do you get them to recognize each other? The word "Lisbon" in one system might be "Lisboa" in another, "940" in a postal code system, or coordinates in a mapping database. Ontology matching is the technology that bridges these differences and it requires exactly the kind of deep linguistic and computational expertise that the Lisbon School has cultivated.

Portugal's longer memory: the scientific tradition that made Lisbon possible

To understand why Lisbon became a center for semantic search research, it helps to understand Portugal's longer scientific history. The broader landscape of science and technology in Portugal is "mainly conducted within a network of research and development units belonging to public universities and state-managed autonomous research institutions." This public infrastructure has been developing for generations.

The historical roots go deep. Pedro Nunes, a mathematician of the Portuguese Renaissance born in 1502, was appointed mathematics teacher at the University of Coimbra in 1537. His post was established specifically "to provide instruction in the technical requirements for navigation" a Portuguese priority at the height of the Age of Discovery, when control of sea trade was the primary source of national wealth. Mathematics became an independent academic post in 1544.

This early institutionalization of mathematics and technical education created a tradition that persisted across centuries. By the 18th century, under the Marquis of Pombal, the University of Coimbra was modernized with the appointment of new professors, both Portuguese and foreign, and the establishment of facilities directed toward teaching natural sciences. The Lisbon Academy of Sciences, one of the oldest learned societies in Portugal, was also founded in the 18th century.

António Egas Moniz, a Portuguese neurologist, won the Nobel Prize in Physiology or Medicine in 1949. His work on cerebral angiography and the development of prefrontal leukotomy represented Portuguese neuroscience at the global frontier.

What this history establishes is a context: Portugal has long maintained institutions, traditions, and expertise in scientific research, even when its resources were modest relative to larger European nations. The country has consistently found ways to contribute to international science and to train researchers who could work at the highest levels.

The Polytechnic University of Lisbon: training the next generation

The Polytechnic University of Lisbon, established in 1986, represents the modern institutional expression of this scientific tradition. The university known until 2026 as the Polytechnic Institute of Lisbon is a public technical university offering bachelor's, master's, and postgraduate degrees in fields including engineering, business, health, education, communication, music, film, and dance.

The university consists of six schools and two higher education institutes spread throughout Lisbon. Among them, the Lisbon School of Engineering (ISEL) stands out as one of the oldest and most respected engineering institutions in the country, founded in 1852. ISEL emphasizes practical, hands-on training alongside theoretical knowledge, fostering innovation and technical expertise. The school is well-regarded for its research initiatives and strong connections with the engineering industry.

The presence of a strong engineering school within the same university system as Lisbon's linguistic research centers creates institutional conditions for the kind of cross-disciplinary work that semantic search demands. When linguists and computer scientists share a city, a public university system, and informal networks of collaboration, the conditions for innovation are present.

The Polytechnic University of Lisbon lists its colors as navy blue and light blue. Its president is Elmano Margato. More than 13,500 students are enrolled. These are the next generation of researchers, engineers, and practitioners people who will carry the Lisbon tradition into new applications and new decades.

What this means for WebSearches readers

For readers researching practitioners, frameworks, and ideas in search and discovery, the Lisbon School offers an important case study in how semantic search capabilities actually develop. This is not a story of a single breakthrough or a single inventor. It is a story of institutional accumulation of decades of linguistic research creating the knowledge base that computational methods later drew upon.

The practical takeaway is this: when evaluating semantic search tools, platforms, or frameworks, the depth of underlying linguistic research matters. The ability to handle multilingual content, to match ontologies across different systems, to reason about meaning beyond just keywords these capabilities have roots in exactly the kind of long-term linguistic scholarship that institutions like CLUL have conducted.

The Lisbon tradition also illustrates the value of interdisciplinary training. CLUL explicitly brings together linguistics, psychology, computer science, and other fields. The Lisbon Semantic Data Lab sits at the intersection of computer science and language research. The most effective semantic search approaches, it appears, are those that take language seriously not as a data format, but as a complex human phenomenon that requires deep study.

For those building search experiences, answer engines, or discovery systems, the Lisbon School suggests looking for partners and platforms with genuine roots in linguistic research, not just statistical pattern matching. The difference often shows up in edge cases: how a system handles polysemy, cultural context, multilingual queries, or ontological complexity. These are exactly the problems that a century of Portuguese linguistics has been working to solve.

The lineage continues

In August 2026, the work that began in a 1932 philology center continues in laboratories across Lisbon. The Lisbon Semantic Data Lab maintains active repositories, with recent commits showing development continuing into late August 2026. CLUL continues to train researchers and publish in language sciences. The Polytechnic University of Lisbon continues to graduate engineers and scientists who carry these traditions forward.

The remarkable thing about this story is its quietness. There are no press releases, no venture capital announcements, no Silicon Valley valuations. There is a center founded in 1932 to study words, a laboratory founded in the 2000s to make those words searchable, and a tradition of scholarship that connects them. This is how real progress in semantic understanding often happens not in dramatic breakthroughs but in decades of patient, careful work by people who believe that understanding language is worth a lifetime.

The Lisbon School did not set out to shape modern semantic search. It set out to understand language. But understanding language, it turns out, is exactly what semantic search requires.

Where to read further

Those interested in exploring the Lisbon School's work directly can start with the Center of Linguistics of the University of Lisbon's institutional overview, which details its history back to 1932 and its current research scope. The Lisbon Semantic Data Lab maintains a public presence on GitHub where its repositories including Exact-OM, KGE_Predictions_GD, and related projects are available for examination. For the broader Portuguese scientific context, the history of science and technology in Portugal provides essential background on the institutional traditions that made Lisbon's linguistic research possible.