Feb. 6, 2024, 5:54 a.m. | Alexandra Saliba Yuanchao Li Ramon Sanabria Catherine Lai

cs.CL updates on arXiv.org arxiv.org

The efficacy of self-supervised speech models has been validated, yet the optimal utilization of their representations remains challenging across diverse tasks. In this study, we delve into Acoustic Word Embeddings (AWEs), a fixed-length feature derived from continuous representations, to explore their advantages in specific tasks. AWEs have previously shown utility in capturing acoustic discriminability. In light of this, we propose measuring layer-wise similarity between AWEs and word embeddings, aiming to further investigate the inherent context within AWEs. Moreover, we evaluate …

advantages analysis continuous cs.cl cs.sd diverse eess.as embeddings emotion explore feature layer recognition specific tasks speech speech emotion study tasks wise word word embeddings

Doctoral Researcher (m/f/div) in Automated Processing of Bioimages

@ Leibniz Institute for Natural Product Research and Infection Biology (Leibniz-HKI) | Jena

Research Scholar (Technical Research)

@ Centre for the Governance of AI | Hybrid; Oxford, UK

Backend Spark Developer

@ Talan | Warsaw, Poland

Pricing & Data Management Intern

@ Novelis | Atlanta, GA, United States

Sr Data Engineer

@ Visa | Bengaluru, India

Customer Analytics / Data Science - Lead Analyst - Analytics US Timezone

@ dentsu international | Bengaluru, India