Latest articles
Beyond Static Chunking: Multimodal and Adaptive Segmentation for Improved Information Retrieval
Introduction Retrieval-Augmented Generation systems [Li et al., 2025], called RAG, rely heavily on how documents are segmented before indexing. This preprocessing step, known as chunking, directly affects retrieval quality and, consequently, the performance of downstream question-answering systems. In practice, chunking strategies [Jain et al., 2025] hardly take into account the document structure; instead, documents are usually split into fixed-size segments without considering their semantic content or modality. Some more advanced
The legal impact of digitalisation of administration on public services
The digital transformation does not only concern the private sector, even though it’s frequently described as more innovative. From computerisation of public administration in the 60’ to the latest applications of artificial intelligence by territorial authorities, public action seems more and more digitalised (and still “in digitalisation”). Therefore, this public action in practice resulting in
🇫🇷 L’IA au service des collectivités locales
https://www.research-bl.com/wp-content/uploads/2019/08/frigg-1-online-video-cutter.com_.mp4 En France, les collectivités se différencient selon plusieurs caractéristiques telles que les budgets qu’elles utilisent au cours d’une année, les investissements qu’elles opèrent et enfin les dettes qu’elles contractent. Le but étant d’apporter une aide aux collectivités locales en développant un modèle capable d’apporter une aide à la prise de décision en leur permettant
🇫🇷 Hermès Démo
CONTEXTE Les documents des marchés publics manifestent une hétérogénéité importante et des similarités trompeuses. Autrement dit, ces documents contiennent généralement des informations typiques comme le nom de l’organisme public, son code SIRET, sa géolocalisation, des dates, les critères et les modalités de choix des candidats, etc. Toutefois, ces informations sont généralement présentées sans aucun format
Blockchain for interoperability
Interoperability is the possibility for different systems to communicate with each other without depending on a particular actor. It is based on the use of an open standard Context The sharing of data between information systems (IS) becomes essential to ensure communication between these IS. This exchange has consequences of compatibility (because of heterogeneous data
Automatic Reconstruction of Sanitation Network
Underground networks are a direct consequence of urbanization. These networks are daily used to offer all the vital services : electricity, Internet, water, etc. However, the available data related to them in particular sanitation ones are various, and come in different types (texts, images, GIS etc.) and formats (analog, digital). In addition, these multisource/multi-format data are
Analysis of user activity (traces) in software
The software for public services are always more complex as the regulation evolves constantly and the user requirement continues to refine. Thus, the quality of the software declines, making it more difficult to maintain and use. For example, some anomalies occurring in the production environment cannot be reproduced and therefore cannot be resolved. From the
🇫🇷 Une analyse automatique du language du “Grand Debat National”
PREAMBULE Cette page propose des résultats d’analyse du corpus du Grand Débat National réalisé au sein de la DRI à Berger-Levrault. L’objectif de cette page est d’illustrer des analyses possibles sur ce type de corpus textuel. Nous nous sommes efforcés d’etre le plus transparent possible dans les techniques effectuées. ACCUEIL Le grand débat est d’intérêt
Automatic software migration: from GWT to Angular
Support for the automation of Web application interface migration: the case of GWT to Angular During the evolution of an application, it is sometimes necessary to migrate its implementation to a different programming language and/or Graphical User Interface (GUI) framework. Web GUI frameworks in particular evolve at a fast pace. For example, in 2018 there
Robustness & Management of Uncertainty in planning
Context The aging population and increasing life expectancy, as we can see today in France and in other developed countries, lead to an increase in the number of the elderly with loss of autonomy and in dangers caused by frailty. In particular, they suffer from chronic diseases with a long-term need for assistance. They largely
Franco-German Dialogue Among AI Industry Leaders (Paris – France)
PMS 2026 – 34th International Conference on Program Comprehension (Toulouse – France)
ICPC – 34th International Conference on Program Comprehension (Rio de Janeiro – Brazil)
CHI 2026 – International conference on Human-Computer Interaction (Barcelona – Spain)
A New Milestone in Building a European AI Ecosystem for Industry
Last Friday marked a significant step forward for the European ecosystem of artificial intelligence applied to industry. After more than a year of close collaboration, the report resulting from the Franco-German dialogue on industrial AI was officially presented to French and German authorities during a dedicated forum held at the Ministry of Economy. This initiative,
Yearbook Research & Innovation 2025: Governed, Responsible, and reality-based Research!
We are proud to announce the publication of our 2025 Research & Innovation Yearbook! More than just an annual review, this document marks a structural shift in the way we design and conduct research. Innovation is no longer seen as a simple technological promise, but as a demanding discipline, confronted with the constraints of reality
Camille Dupré Ph.D. thesis defense: “Pad-based Interaction in Mixed Reality environments”
Thursday 18th December at 2p.m. Paris time, Camille Dupré, Ph.D. Candidate has defended her thesis named “Pad-based Interaction in Mixed Reality environments”. Her thesis defense took place at the LISN, in Gif-sur-Yvette (660 Av. des Sciences Bâtiment, 91190), France. Take a look at the summary below. Summary Mixed Reality (MR) environments integrate virtual elements into
Nihed Bendahman Ph.D. thesis defense: “Evaluation and mitigation of hallucinations in automatic summarization in the specific context of legal documents”
Monday 15th December at 2p.m. Paris time, Nihed Bendahman, Ph.D. Candidate has defended her thesis named “Evaluation and mitigation of hallucinations in automatic summarization in the specific context of legal documents”. Her thesis defense took place at the IRIT Research Laboratory, in Toulouse, France. Take a look at the summary below. Summary Legal monitoring is a
Gabriel Darbord Ph.D. thesis defense: “Automatic test generation to help modernize our applications”
Friday 5th December at 9a.m. Paris time, Gabriel Darbord, Ph.D. Candidate has defended his thesis named “Automatic test generation to help modernize our applications”. His thesis defense took place in Lille, France. Take a look at the summary below. This thesis is fully in line with the partnership between Berger-Levrault and Inria, which aims to
Berger-Levrault strengthens its ties with AI startups!
We are proud to announce that we have joined Hub France IA, the largest association dedicated to artificial intelligence in France. This network now brings together more than 250 members—companies, startups, research laboratories, and institutions—who share the same goal: to accelerate the development and adoption of AI in France and Europe. Getting closer to the
Celebrating New PhDs from the BL.Research Team!
At Berger-Levrault, research is more than a mission—it’s a shared adventure. As the new academic year begins, we are proud to celebrate the success of four of our colleagues from the BL.Research team, who have reached a major milestone in their scientific journeys: the defense of their doctoral theses. These achievements are the result of
Hamza Safri Ph.D. thesis defense: “Federated learning for the IoT : Application for Industry 4.0”
Thuesday 24th June at 3pm Paris time, Hamza Safri, Ph.D. Candidate has defended his thesis named “Federated learning for the IoT: Application for Industry 4.0”. His thesis defense took place at the Inria Minatec Grenoble, Grenoble, France. Take a look at the summary below. Keywords: Model generalization, predictive maintenance, industrial IoT, federated learning, edge network,
