Projects

Members of the lab take an active part in European research projects, as coordinators and as partners.

01 Current

Funded projects

Horizon Europe funds the research, the people and the collaborations listed here.

Active ARMADA

ARMADA Reliable Conversational Domain-specific Data Exploration and Analysis

ARMADA is a doctoral network training fifteen early-stage researchers on what it takes for a large language model to be reliable when it assists analysis in a sensitive domain. A reliable model, in the project’s formulation, produces answers that are timely, consistent and verifiable.

The training programme is organised around several axes: alignment with domain knowledge, explainability and soundness of the answers, reactivity through interactive correction workflows, and the effectiveness and efficiency of the underlying computation. The researchers work across natural language processing, machine learning, data management and algorithms, and the network is designed so that its graduates can advise European bodies on the regulation of AI in critical sectors.

The lab’s researchers work on the data-management side of that programme: grounding an answer in data that can be inspected, and giving a person a way to intervene when the answer is wrong.

Related research themes: Human-centered analytics, Validated and trustworthy data systems.

Active DataGEMS

DataGEMS Data Discovery Platform with Generalized Exploratory, Management, and Search Capabilities

DataGEMS is a next-generation dataset discovery and management ecosystem. It takes in data of different kinds (structured and unstructured, real-time and historical) and enriches it through data profiling so that it can be found at all.

On top of that, the platform lets people discover and analyse data both across and within datasets, using mechanisms meant for humans rather than for query engines: natural language, patterns, and stepwise guidance while datasets are being combined. The whole thing is built on the principles of data FAIRness, openness and re-use, and is delivered as a service in the European Open Science Cloud Exchange.

Its stated aim is to bridge the gap between the people who publish data and the people who need it. That is the same gap the lab’s researchers work on from the exploration side.

02 Papers and code

Papers and code

The lab’s peer-reviewed papers are on the publications page, each one tagged with the project that funded it, where one did. Where a paper has a dataset or code of its own, its entry links to them. Code is in two places: a project’s own organisation, linked from its record above, and the lab’s own repositories for work outside the projects.

Publications Open-source code

Open by default. A paper written under one of these grants names the grant and the agency that funds it, and follows the open science terms that come with it: the accepted version is deposited where it can be read without payment, and the data behind it is managed under the FAIR principles. The pledge to open science on the publications page sets out the rest, including what AI tools may and may not contribute.