EMODE
An emotion-aware spoken language model based on the Dynamic Paralinguistic and Semantic Expert (DPSE) architecture. It decouples speech features into a semantic pathway and a paralinguistic pathway, achieving independent processing and joint optimization of emotion and semantics through dynamic routing and fusion.
Timeline (last 90 days)
EMODE is an emotion-aware spoken language model built around a Dynamic Paralinguistic Specialist Expert (DPSE).
EMODE adopts a three-stage curriculum training approach: semantic warm-up, paralinguistic activation, and joint refinement.
The research team constructed a new bilingual MEPA benchmark for evaluating cross-lingual emotion understanding capabilities.
EMODE achieves a better balance between lexical fidelity and emotional sensitivity on speech emotion recognition tests, empathetic response evaluation, and the MEPA benchmark.
The design principle proposed by EMODE is that emotional information deserves to be explicitly modeled rather than being passively recovered as a by-product of semantics.
All Facts (5)
EMODE is an emotion-aware spoken language model built around a Dynamic Paralinguistic Specialist Expert (DPSE).
50%UnverifiedEMODE adopts a three-stage curriculum training approach: semantic warm-up, paralinguistic activation, and joint refinement.
50%UnverifiedThe research team constructed a new bilingual MEPA benchmark for evaluating cross-lingual emotion understanding capabilities.
50%UnverifiedEMODE achieves a better balance between lexical fidelity and emotional sensitivity on speech emotion recognition tests, empathetic response evaluation, and the MEPA benchmark.
50%UnverifiedThe design principle proposed by EMODE is that emotional information deserves to be explicitly modeled rather than being passively recovered as a by-product of semantics.
50%