Search

Your search keyword '"Li, Danni"' showing total 1,021 results
Start Over Author "Li, Danni"
1,021 results on '"Li, Danni"'

5. TODM: Train Once Deploy Many Efficient Supernet-Based RNN-T Compression For On-device ASR Models

Author

Shangguan, Yuan, Yang, Haichuan, Li, Danni, Wu, Chunyang, Fathullah, Yassir, Wang, Dilin, Dalmia, Ayushi, Krishnamoorthi, Raghuraman, Kalinli, Ozlem, Jia, Junteng, Mahadeokar, Jay, Lei, Xin, Seltzer, Mike, and Chandra, Vikas

Subjects
Computer Science - Computation and Language, Computer Science - Machine Learning, Computer Science - Sound, Electrical Engineering and Systems Science - Audio and Speech Processing
Abstract

Automatic Speech Recognition (ASR) models need to be optimized for specific hardware before they can be deployed on devices. This can be done by tuning the model's hyperparameters or exploring variations in its architecture. Re-training and re-validating models after making these changes can be a resource-intensive task. This paper presents TODM (Train Once Deploy Many), a new approach to efficiently train many sizes of hardware-friendly on-device ASR models with comparable GPU-hours to that of a single training job. TODM leverages insights from prior work on Supernet, where Recurrent Neural Network Transducer (RNN-T) models share weights within a Supernet. It reduces layer sizes and widths of the Supernet to obtain subnetworks, making them smaller models suitable for all hardware types. We introduce a novel combination of three techniques to improve the outcomes of the TODM Supernet: adaptive dropouts, an in-place Alpha-divergence knowledge distillation, and the use of ScaledAdam optimizer. We validate our approach by comparing Supernet-trained versus individually tuned Multi-Head State Space Model (MH-SSM) RNN-T using LibriSpeech. Results demonstrate that our TODM Supernet either matches or surpasses the performance of manually tuned models by up to a relative of 3% better in word error rate (WER), while efficiently keeping the cost of training many models at a small constant., Comment: Meta AI; Submitted to ICASSP 2024

Published
2023

19. Retrospective systematic analysis of incidence and severity of COVID-19 reinfection

Author

SHI Lei, XIN Shen, SUN Qin, WANG Yuan, LI Danni, and LI Ke

Subjects
covid-19, reinfection, critical reinfection, Medicine
Abstract

ObjectiveTo investigate the incidence and severity of COVID-19 reinfection.MethodsWe searched relevant studies on COVID-19 reinfection， including cohort study， case report， and cross-sectional study in PubMed， Embase， CNKI and Wanfang databases. Revman5.3.0 was used for statistical analysis.ResultsA total of 52 studies in 19 countries were included. These literatures showed moderate and high quality. Furthermore， the pooled incidence of COVID-19 reinfection was estimated to be 1.9%（95%CI： 0.9%‒4.7%， P

Published
2023
Full Text
View/download PDF

21. 'They're Going to Forget about Their Mother Tongue': Influence of Chinese Beliefs in Child Home Language and Literacy Development

Author

Gonzalez, Jorge E., Liew, Jeffrey, Zou, Yali, Curtis, Gayle, and Li, Danni

Abstract

The primary aim of this study was to explore and understand Chinese American parents' language and literacy beliefs, perceptions, and practices with regard to their preschool-age children. Using a convergent parallel mixed-methods study design, parents (seven mothers and one father) were recruited from a community center in a large southern city in the United States to participate in focus groups. Focus group transcripts were analyzed using a side-by-side approach to compare qualitative data from discussions in the focus groups with quantitative data from parents' responses to a questionnaire. Six themes emerged from the parent focus group data: (a) expectation of bilingual/trilingual child language ability and concern about the loss of heritage language; (b) strategies to support child learning and language development; (c) learning-related tools and resources; (d) parent expectations of schools with regard to child learning and literacy development; (e) preparing children for a competitive future; and (f) challenges with supporting child language growth and literacy development. Parents were very clear about their roles regarding their child's language development, with mothers and fathers taking on different roles depending on the learning situation and language in use. The quantitative data supported the qualitative findings. Directions for future research are suggested.

Published
2022
Full Text
View/download PDF

22. Cultural Values, Intergenerational Transmission of Internalized Racism, Education, and Career Goals in Chinese American Families

Author

Hill, Daphne J., Li, Danni, Wang, Jun, and Liew, Jeffrey

Abstract

The present study is the first of its kind using a dyadic and prospective research design to test whether traditional Asian cultural values and internalized racism among Chinese American adolescents and their first-generation immigrant parents are influential factors in the educational and vocational goals of the adolescents by the time they become young adults. While results show that traditional Asian cultural values and internalized racism were not related to adolescents' educational aspiration or young adults' major choice, results do show an intergenerational correlation in internalized racism. Furthermore, results show that youth who planned to pursue science/health professions had significantly higher levels of internalized racism than those who planned to pursue business/law professions. Findings suggest that internalized racism may restrict parents' expectations and goals for their children and, in tandem, serve as a barrier for youths' self-determination or sense of autonomy in their selection of college majors and career development.

Published
2022
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results