Tootfinder

Towards Effective and Efficient Non-autoregressive Decoding Using Block-based Attention Mask
Tianzi Wang, Xurong Xie, Zhaoqing Li, Shoukang Hu, Zengrui Jing, Jiajun Deng, Mingyu Cui, Shujie Hu, Mengzhe Geng, Guinan Li, Helen Meng, Xunying Liu
https://arxiv.org/abs/2406.10034

Towards Effective and Efficient Non-autoregressive Decoding Using Block-based Attention Mask
This paper proposes a novel non-autoregressive (NAR) block-based Attention Mask Decoder (AMD) that flexibly balances performance-efficiency trade-offs for Conformer ASR systems. AMD performs parallel NAR inference within contiguous blocks of output labels that are concealed using attention masks, while conducting left-to-right AR prediction and history context amalgamation between blocks. A beam search algorithm is designed to leverage a dynamic fusion of CTC, AR Decoder, and AMD probabilities.…

The Japanese Vision for the Black Hole Explorer Mission
Kazunori Akiyama, Kotaro Niinuma, Kazuhiro Hada, Akihiro Doi, Yoshiaki Hagiwara, Aya E. Higuchi, Mareki Honma, Tomohisa Kawashima, Dimitar Kolev, Shoko Koyama, Sho Masui, Ken Ohsuga, Hidetoshi Sano, Hideki Takami, Yuh Tsunetoe, Yoshinori Uzawa, Takuya Akahori, Yuto Akiyama, Peter Galison, Takayuki J. Hayashi, Tomoya Hirota, Makoto Inoue, Yuhei Iwata, Michael D. Johnson, Motoki Kino, Yutaro Kofuji, Yosuke Mizuno, Kotaro Moriyama, H…

The Japanese Vision for the Black Hole Explorer Mission
The Black Hole Explorer (BHEX) is a next-generation space very long baseline interferometry (VLBI) mission concept that will extend the ground-based millimeter/submillimeter arrays into space. The mission, closely aligned with the science priorities of the Japanese VLBI community, involves an active engagement of this community in the development of the mission, resulting in the formation of the Black Hole Explorer Japan Consortium. Here we present the current Japanese vision for the mission, r…

Deciphering the imprint of AGN feedback in Seyfert galaxies: Nuclear-scale molecular gas deficits
S. Garc\'ia-Burillo, E. K. S. Hicks, A. Alonso-Herrero, M. Pereira-Santaella, A. Usero, M. Querejeta, O. Gonz\'alez-Martin, D. Delaney, C. Ramos Almeida, F. Combes, D. Angl\'es-Alc\'azar, A. Audibert, E. Bellocchi, R. I. Davies, T. A. Davis, J. S. Elford, I. Garc\'ia-Bernete, S. H\"onig, A. Labiano, M. T. Leist, N. A. Levenson, E. L\'opez-Rodr\'iguez, J. Me…

Deciphering the imprint of AGN feedback in Seyfert galaxies: Nuclear-scale molecular gas deficits
We use a sample of 64 nearby (D=7-45 Mpc) disk galaxies including 45 AGN and 19 non-AGN, that have high spatial resolution multiline CO observations obtained with the ALMA and/or PdBI arrays to study the distribution of cold molecular gas in their circumunuclear disks (CND). We analyze whether the concentration of cold molecular gas changes as a function of the X-ray luminosity in the 2-10 keV range ($L_{\rm X}$). We also study the concentration of the hot molecular gas using NIR data obtained …

Inclusive ASR for Disfluent Speech: Cascaded Large-Scale Self-Supervised Learning with Targeted Fine-Tuning and Data Augmentation
Dena Mujtaba, Nihar R. Mahapatra, Megan Arney, J. Scott Yaruss, Caryn Herring, Jia Bin
https://arxiv.org/abs/2406.10177

Inclusive ASR for Disfluent Speech: Cascaded Large-Scale Self-Supervised Learning with Targeted Fine-Tuning and Data Augmentation
Automatic speech recognition (ASR) systems often falter while processing stuttering-related disfluencies -- such as involuntary blocks and word repetitions -- yielding inaccurate transcripts. A critical barrier to progress is the scarcity of large, annotated disfluent speech datasets. Therefore, we present an inclusive ASR design approach, leveraging large-scale self-supervised learning on standard speech followed by targeted fine-tuning and data augmentation on a smaller, curated dataset of di…