While speaker adaptation for end-to-end speech synthesis using speaker embeddings can produce good speaker similarity for speakers seen during training, there remains a gap for zero-shot adaptation to unseen speakers. We investigate multi-speaker modeling for end-to-end text-to-speech synthesis and study the effects of different types of state-of-the-art neural speaker embeddings on speaker similarity for unseen speakers. Learnable dictionary encoding-based speaker embeddings with angular softmax loss can improve equal error rates over x-vectors in a speaker verification task; these embeddings also improve speaker similarity and naturalness for unseen speakers when used for zero-shot adaptation to new speakers in endto-end speech synthesis.

Automatic methods to predict listener opinions of synthesized speech remain elusive since listeners, systems being evaluated, characteristics of the speech, and even the instructions given and the rating scale all vary from test to test. While automatic predictors for metrics such as mean opinion score (MOS) can achieve high prediction accuracy on samples from the same test, they typically fail to generalize well to new listening test contexts. In this paper, using a variety of networks for MOS prediction including MOSNet and selfsupervised speech models such as wav2vec2, we investigate their performance on data from different listening tests in both zero-shot and fine-tuned settings. We find that wav2vec2 models fine-tuned for MOS prediction have good generalization capability to out-ofdomain data even for the most challenging case of utterance-level predictions in the zero-shot setting, and that fine-tuning to in-domain data can improve predictions. We also observe that unseen systems are especially challenging for MOS prediction models.

The spoken term detection (STD) task aims to return relevant segments from a spoken archive that contain the query terms whether or not they are in the system vocabulary. This paper focuses on pronunciation modeling for Out-of-Vocabulary (OOV) terms which frequently occur in STD queries. The STD system described in this paper indexes word-level and sub-word level lattices or confusion networks produced by an LVCSR system using Weighted Finite State Transducers (WFST). We investigate the inclusion of n-best pronunciation variants for OOV terms (obtained from letter-to-sound rules) into the search and present the results obtained by indexing confusion networks as well as lattices. The following observations are worth mentioning: phone indexes generated from sub-words represent OOVs well and too many variants for the OOV terms degrade performance if pronunciations are not weighted.

No abstract

Can we identify which utterances in a corpus are the best to use for voice training, based on acoustic/prosodic features, and which utterances should be excluded because they will introduce noise, artifacts, or inconsistency into the voice? Can we use found data such as radio broadcast news to build HMM-based synthesized voices? Can we select a subset of training utterances from a corpus of found data to produce a better voice than one trained on all of the data? Which voice training and modeling approaches work best for this type of data? Data and Tools Boston University Radio News Corpus (BURNC): 7+ hours of professionally-read radio broadcast news from 3 female and 4 male speakers Challenges: Multiple speakers, non-TTS speaking style