PhD topic: Hybrid Evaluation of Natural Language Queries over Heterogeneous Data Lakes

Dear all, We are opening a funded PhD topic at LISN, Université Paris-Saclay, on Hybrid Evaluation of Natural Language Queries over Heterogeneous Data Lakes. The full PhD proposal is available here: https://phparis.net/uploads/topol_phd.pdf The PhD will be carried out in the LaHDAK team, in the context of the TopOL project. The general objective is to design methods for answering natural language questions over heterogeneous data lakes combining structured and unstructured sources: relational databases, CSV/JSON/XML files, RDF and property graphs, PDF and text documents, Office documents, open data, and external knowledge sources. The thesis will investigate how to combine: - natural language query interpretation, possibly using AMR-based graph representations; - graph-based query evaluation over entities and relationships; - vector-based retrieval and RAG over unstructured data; - reconciliation of symbolic and neural retrieval results; - provenance-aware and explainable answers; - estimation of the temporal validity of query answers. The application context is data exploration for non-technical users, with a particular focus on investigative journalism, where traceability, confidentiality, and reliability are essential. The candidate should have a strong background in computer science and an interest in one or several of the following areas: data management, knowledge graphs, semantic web technologies, information retrieval, natural language processing, retrieval-augmented generation, knowledge representation, or explainable AI. Location: LISN, Université Paris-Saclay, France Doctoral school: STIC, Université Paris-Saclay Duration: 36 months Expected start: 2026 Supervision: Fatiha Saïs, Pierre-Henri Paris, and Benoît Groz Interested candidates may contact using this address: topol@lisn.upsaclay.fr