JUMAS addresses the need to build an infrastructure able to optimise the information workflow in order to facilitate later analysis. New models and techniques for representing and automatically extracting the embedded semantics derived from multiple data sources will be developed. The most important goal of the JUMAS system is to collect, enrich and share multimedia documents annotated with embedded semantic minimising manual transcription activity. JUMAS is tailored at managing situations in which multiple cameras and audio sources are used to record assemblies in which people debates and event sequences need to be semantically reconstructed for future consultations. The prototype of JUMAS will be tested interworking with legacy systems, but the system can be viewed as able to support business processes and problem-solving in a variety of domains.
Our pick of the week by @FBKZhihangXie: "#Speech Discrete Tokens or Continuous Features? A Comparative Analysis for Spoken Language Understanding in #SpeechLLMs" by @WangDingdo2603, Junan Li, @HelenMeng_CUHK, et al. (#EMNLP2025)
#SLU #SpeechTech
๐ New paper: Speech Discrete Tokens or Continuous Features?
๐ https://aclanthology.org/2025.emnlp-main.1266.pdf
๐งฉ A comprehensive benchmark of SpeechLLMs using HuBERT/WavLM with Qwen & LLaMA.
โจ Continuous features outperform overall, while discrete tokens excel at phoneme-level detail.
๐ Exciting news from the @FBK_MT group!
Four of our members @BeatriceSavoldi, @lina_conti, @negri_teo & @luisabentivogli are attending #EMNLP2025 in Suzhou ๐จ๐ณ with 5 accepted papers!
Come to our sessions & let's connect:
๐ https://mt.fbk.eu/fbk-mt-at-emnlp-2025/
Weโre also hiring postdocs!โก
๐๐Congratulations to our PhD student @DennisFucci on a very successful thesis defense! ๐
Many thanks to the evaluation committee members @debora_nozza, @mirco_ravanelli, and Leonardo Badino for their insightful feedback and appreciation of his work!
#nlproc