In recent years the field of research on natural language interfaces for databases (NLIDBs) has progressed considerably, which can be seen by the results of challenges like Spider. However, most of these approaches concentrate on delivering best-guess answers and improving (computational) accuracy.
Yet, there are still a lot of open issues regarding robustness, confidence, and transparency. Therefore, this vision paper starts to point to relevant milestones as well as corresponding opportunities for addressing them, opening up a potential path of the future development of NLIDBs.
- Christoph Brandt (TU Darmstadt)
- Benjamin Hättasch (TU Darmstadt)
- Nadja Geisler (TU Darmstadt)
- Carsten Binnig (TU Darmstadt)