There are two interleaved high-level goals for the EXCITEMENT project. The first is to set up, for the first time, a generic architecture and a comprehensive implementation for a multilingual textual inference platform and to make it available to the scientific and technological communities. The second goal of the project is to develop a new generation of inference-based industrial text exploration applications for customer interactions, which enables businesses to better analyze and make sense of their diverse and often unpredicted client content. These goals were be achieved for three languages, i.e. English, German and Italian, and for three customer interaction channels, i.e. speech (transcriptions), email and social media.
Our pick of the week by
@lina_conti
: "Greater accessibility can amplify discrimination in generative AI" by
@CarolinHolterm, @minhducbui_nlp, @KaitlynZhou, @vjhofmann, @kelina1124, @anne_lauscher
📰
#GenderBias #SpeechLLM
Pick of the week @fbk_mt: "Greater accessibility can amplify discrimination in generative AI"
Gender bias in speech-based LLMs examined from multiple angles: a user survey, automatic bias measurement, and pitch manipulation experiments.
https://arxiv.org/pdf/2603.22260
Late update, but we had two great talks last month!
#MachineTranslation #FBK #NLProc #GenderBias #SpeechSynthesis
Our pick of the week by @dhairya_su47605
: "Scaling Laws for Precision" by @tanishqkumar07, Zachary Ankner, @bfspectorShiekh, @blake__bordelon, @Muennighoff, @mansiege, @CPehlevan, Christopher R´e, @AdtRaghunathan
📰
#Quantization #LLM #ScalingLaw
Pick of the week @fbk_mt
Super interesting paper on the limitations of quantization, demonstrating how post-training quantization scales poorly in data.
https://arxiv.org/abs/2411.04330
⭐ For our #PickOfTheWeek, this paper explores an important question for modern speech AI:
🎙️ Which Evaluation for Which Speech Model?
👥 Authors: @Maureendss , @EeshanDhekane
Speech foundation models are evolving rapidly, but evaluation practices are still fragmented.