Project Details
Multilingual and Cross-cultural interactions for context-aware, and bias-controlled dialogue systems for safety-critical applications
Project Period: 1. 1. 2024 – 31. 12. 2026
Project Type: grant
Code: SEP-210943216
Agency: Evropská unie
Program: HORIZON EUROPE
Human computer interaction and interface, visualization and natural language, artificial intelligence, intelligence systems, multi agents systems, natural language processing, data protection and privacy, machine learning, statistical data processing and applications using data processing, formal, cognitive, functional and computational linguistics, distributed and federated adaptation of Large Language Models, Multilinguality, Multimodality, Human-in-the-loop, Bias mitigation, Grounding.
ELOQUENCE aims to research and develop new technologies supporting collaborative voice/chat bots for both low secure (low risk) and highly secure (high risk) applications. Dialogue engines powered by voice assistants have already been present in various commercial/governmental applications with lower or higher level complexities. In both cases, this complexity can be translated to a problem of analysing unstructured dialogues. Key objective of ELOQUENCE is to understand unstructured dialogues and conduct them in an explainable, safe, knowledge-grounded, trustworthy and unbiased way, while considering and building on top of prior achievements in this domain (e.g. recently launched chatGPT Large Language Models (LLMs). While including key industrial enterprises from Europe in this project (i.e. Omilia, Telefonica. ...) will approach safety with human-in-the-loop for safety-critical applications (i.e., emergency services) and via information retrieval and fact-checking against an online knowledge base for less critical autonomous systems (i.e., home-assistants). ELOQUENCE will target the R&D of these novel conversational AI technologies in multilingual and multimodal environments. Both basic research and its direct deployment through two pilots will be targeted: 1) emergency call contact centres and 2) smart assistants through decentralised training in smart homes.
Beneš Karel, Ing. (DCGM)
Burget Lukáš, doc. Ing., Ph.D. (DCGM)
Kesiraju Santosh, Ph.D. (DCGM)
Schwarz Petr, Ing., Ph.D. (DCGM)
2024
- YUSUF, B.; ČERNOCKÝ, J.; SARAÇLAR, M. Pretraining End-to-End Keyword Search with Automatically Discovered Acoustic Units. Proceedings of Interspeech 2024. Proceedings of Interspeech. Kos: International Speech Communication Association, 2024.
p. 5068-5072. ISSN: 1990-9772. Detail