We propose autoencoding speaker conversion for training data augmentation in automatic speech translation. This technique directly transforms an audio sequence, resulting in audio synthesized to resemble another speaker's voice. Our method compares favorably to SpecAugment on English-French and English-Romanian automatic speech translation (AST) tasks as well as on a low-resource English automatic speech recognition (ASR) task. Further, in ablations, we show the benefits of both quantity and diversity in augmented data. Finally, we show that we can combine our approach with augmentation by machine-translated transcripts to obtain a competitive end-toend AST model that outperforms a very strong cascade model on an English-French AST task. Our method is sufficiently general that it can be applied to other speech generation and analysis tasks.

Current preimplantation genetic testing (PGT) enables the selection of embryos based on fetal aneuploidy or the presence a small number of preselected disease-associated variants. Here we present a new approach that takes advantage of the improved genome coverage and uniformity of primary template-directed amplification (PTA) to call most early embryo genetic variants accurately and reproducibly from a preimplantation biopsy. With this approach, we identified clonal and mosaic chromosomal aneuploidy, de novo mitochondrial variants, and variants predicted to cause mendelian and non-mendelian diseases. In addition, we utilized the genome-wide information to compute polygenic risk scores for common diseases. Although numerous computational, interpretive, and ethical challenges, this approach establishes the technical feasibility of screening for and preventing numerous debilitating inherited diseases.

No abstract

No abstract