Work

Survey on Guided Natural Language Interface to Databases in the era of LLM

Year: 2026

Type: article

Abstract: Deliverable D5.1 (WP5 midterm state-of-the-art report) of the ARMADA MSCA Doctoral Network, funded by Horizon Europe under Grant Agreement No. 101168951. Natural language interfaces to databases (NLIDBs) aim to answer users' natural language queries over databases on structured data. In this survey, we take a broad view of the interfaces to databases including relational databases and knowledge graphs, whose data types can be structured or hybrid data with text column. We design a three-stage framework as an analytical lens for organizing existing studies. (i)Query reformulation in request interpretation stage, (ii)Semantic parsing in program construction stage, (iii)Plan optimization in program execution stage. We review representative systems and benchmarks from the database and NLP communities and highlight open problems in faithfulness, evaluation reliability, cross-formalism transfer, and deployable efficiency.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.23620362

Authors: Dong Yifei, Babatunde, Ibukun David, Jiang, Yuxuan, Velegrakis, Yannis, Koutrika, Georgia, Gionis, Aristides, Hose, Katja, Mottin, Davide

Institutions: Wuhan University, Université Mohammed VI Polytechnique, South China University of Technology, University of Trento, Utrecht University, KTH Royal Institute of Technology, TU Wien, Aalborg University, Aarhus University

Language: English

Topic: Natural Language Processing Techniques

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

Funder: European Commission

OpenAlex ID: W7223447136

FWCI:

Cites:

Cited by:

Related to: 10