We are specialized in human-machine vocal interaction systems (dialogue systems, human-robot situated interactions). Some keywords to define our recent endeavours: reinforcement learning for dialogue policies, online learning, social competencies, humor traits, zero-shot semantic parser. More at https://cv.archives-ouvertes.fr/fabricelefevre

We believe HMI can be a true asset to sustain AI explanabiility by helping AI solutions to communicate with users their ways and motives. Also, at the internal level, such systems can be improved to better disclose verbally their own limits during the interactions. In our researches we use both architectures for our systems: pipeline of modules (ASR, SLU, DSTC, DM, NLG, TTS), old fashion but efficient in practice, and end-to-end neural networks, trendy and promising but still brittle.

At Avignon University LIA lia.univ-avignon.fr is the Computer Science laboratory. It has a strong expertise in all the fields of NLP, especially in speech recognition and understanding, meta-data extraction, diarization, interactive systems.