Eric Battenberg, RJ Skerry-Ryan, Daisy Stanton, Soroosh Mariooryad, Matt Shannon, Julian Salazar, David Teh-Hwa Kao. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

This work explores the task of synthesizing speech in non-existent human-sounding voices. We call this task "speaker generation", and present TacoSpawn, a system that performs competitively at this task. TacoSpawn is a recurrent attention-based text-to-speech model that learns a distribution over a speaker embedding space, which enables sampling of novel and diverse speakers. Our method is easy to implement, and does not require transfer learning from speaker ID systems. We present objective and subjective metrics for evaluating performance on this task, and demonstrate that our proposed objective metrics correlate with human perception of speaker similarity. Audio samples are available on our demo page1.

The recent development of data-driven AI promises to automate medical diagnosis; however, most AI functions as 'black boxes' to physicians with limited computational knowledge. Using medical imaging as a point of departure, we conducted three iterations of design activities to formulate CheXplain — a system that enables physicians to explore and understand AI-enabled chest X-ray analysis: (i) a paired survey between referring physicians and radiologists reveals whether, when, and what kinds of explanations are needed; (ii) a low-fidelity prototype co-designed with three physicians formulates eight key features; and (iii) a high-fidelity prototype evaluated by another six physicians provides detailed summative insights on how each feature enables the exploration and understanding of AI. We summarize by discussing recommendations for future work to design and implement explainable medical AI systems that encompass four recurring themes: motivation, constraint, explanation, and justification.

Despite the ability to produce human-level speech for in-domain text, attention-based end-to-end text-to-speech (TTS) systems suffer from text alignment failures that increase in frequency for out-of-domain text. We show that these failures can be addressed using simple location-relative attention mechanisms that do away with content-based query/key comparisons. We compare two families of attention mechanisms: location-relative GMM-based mechanisms and additive energy-based mechanisms. We suggest simple modifications to GMM-based attention that allow it to align quickly and consistently during training, and introduce a new location-relative attention mechanism to the additive energy-based family, called Dynamic Convolution Attention (DCA). We compare the various mechanisms in terms of alignment speed and consistency during training, naturalness, and ability to generalize to long utterances, and conclude that GMM attention and DCA can generalize to very long utterances, while preserving naturalness for shorter, in-domain utterances.