Current sign language generation relies on technical metrics that often overlook capturing actual Deaf user comprehension. This is particularly challenging for Chinese Natural Sign Language (CNSL) with its scene-dependent expressions and spatial grammar. We assembled a scene-aware CNSL generation prototype using established components to serve as a controlled evaluation stimulus. In partnership with four Deaf co-researchers, we developed a seven-dimensional evaluation framework combining objective comprehension tests with validated subjective measures. An evaluation with 24 Deaf participants revealed that while technical metrics indicated success, objective comprehension tests showed lower comprehension and higher cognitive load, particularly for complex spatial grammar. These results demonstrate that current evaluation methods do not align with user needs. Our framework shifts from computer benchmarking toward human-centered assessment, prioritizing comprehension, cognitive load, and cultural authenticity. The findings underscore the importance of creating accessibility technology with, rather than for, Deaf communities.

Speech emotion recognition (SER) is a challenging task whose performance heavily relies on suitable affect-salient representations. Recently, transformer has exhibited outstanding qualities in learning relevant representations associated with this task. However, a normal transformer is only able to process the uni-source input, and there is often only one kind of input feature in a transformer-based SER system, which may cause limited knowledge. In this paper, we attempt to use the cross-attention transformer (CAT) to handle bi-source input. We propose a novel SER system to better fuse three types of acoustic features – raw waveform data, spectrogram, and MFCC using CAT. Experiments conducted on the IEMOCAP benchmark dataset have shown that our proposed system can achieve a 73.80% weighted accuracy (WA) and 74.25% unweighted accuracy (UA), which outperforms existing state-of-the-art approaches.