NESPOLE! System has been developed using two scenarios: the tourism scenario and the first aid medical assistance scenario. During the project life three main data collection have been carried on in order to develop the first and the second showcase. During the first year 191 dialogues have been collected. There are 62 German dialogues recorded, 61 Italian, 37 English and 31 French. Particularly an amount of 6 hours of dialogues for Italian and French, 7 hours for English, 8 hours for German has been recorded. Dialogues were about five predefined tourism scenarios. During the last year two major data collections have been carried on: the first one aimed at expanding the tourism scenario and the second one at addressing the medical domain. For the monolingual data collection five tourism scenarios were developed; 66 dialogues were recorded yielding 994.57 minutes of data: 243.52 minutes comprised in sixteen English dialogues, 246 minutes in sixteen German dialogues, 272.52 minutes in seventeen French dialogues and 232.53 minutes in seventeen Italian dialogues. The data collection on the medical domain involved Italian, English and German languages. A total of 49 dialogues were collected. The recording results in a total of 8 hours 25 minutes of audio files.
Our pick of the week by @mgaido91: "WhisperKit: On-device Real-time ASR with Billion-Scale Transformers" by Atila Orhon, Arda Okan, Berkin Durmus, @zachnagengast, and Eduardo Pacheco (ICML 2025)
#speech #speechtech #whisper #ASR #realtime
A couple of weeks before presenting our large-scale speech model compression task at IWSLT, here there is of the first attempts to bring large-scale models to the devices on the edge: https://arxiv.org/pdf/2507.10860... Hope to see more works along this direction!
Our pick of the week by @FBKZhihangXie: "Adversarial Speech-Text Pre-Training for Speech Translation" by Chenxuan Liu, Liping Chen, Weitai Zhang, Xiaoxi Li, Peiwang Tang, Mingjia Yu, Sreyan Ghosh, and Zhongyi Ye (ICASSP 2025)
#speech #speechprocessing #speechtech #translation
🚀 AdvST: Adversarial training aligns speech and text distributions without parallel data! Combines adversarial learning + hidden-state swapping to fix length mismatch & boost low-resource speech translation. https://ieeexplore.ieee.org/document/10888294
A special evening in Rome to talk about Physical AI and Europe’s role in shaping this new frontier.
Partners from across Europe came together to present the DVPS project, and connect with key people from public institutions, embassies, industries, national & international media.
Thrilled to be part of this amazing project and team!
🚀 DVPS has launched at Translated's HQ!
70 researchers from 20 institutions across 9 countries unite to build next-gen multimodal foundation models that learn from real-world interaction.
A new European AI journey begins.
#DVPS #PhysicalAI #HorizonEurope #MultimodalAI