Error correction is an important post-processing method in speech recognition that aims to detect and correct errors in speech recognition results, thereby further improving the accuracy of speech recognition. Many error correction models use autoregressive models with high latency, but…
The DuSable Museum of African American History (opens in new tab) is a Chicago community institution and the first non-profit Museum dedicated to the collection, documentation, preservation, study and the dissemination of the history and culture of Africans and African…
Just Tech: Centering Community-Driven Innovation at the Margins Episode 3 with Dr. Sasha Costanza-Chock
In “Just Tech: Centering Community-Driven Innovation at the Margins,” Senior Principal Researcher Mary L. Gray explores how technology and community intertwine and the role technology can play in supporting community-driven innovation and community-based organizations. Dr. Gray and her team are…
Awards | 2022 Mining Software Repositories (MSR) conference
Dongmei Zhang awarded the 2022 Mining Software Repositories Foundational Contribution Award
Dongmei Zhang along with fellow Mining Software Repositories (MSR) Conference attendee, Tao Xie, were awarded the 2022 MSR Foundational Contribution award for their fundamental contributions in opening new areas of software analytics research with pioneering results and techniques that have…
STEDII Properties of a Good Metric
Good metrics enable good decisions. What makes a metric good? In this blog post we introduce the STEDII (Sensitivity, Trustworthiness, Efficiency, Debuggability, Interpretability, and Inclusivity) framework to define and evaluate the good properties of a metric and of an A/B…
In the news | The Atlantic
This Is What Happens When There Are Too Many Meetings
After two years of working from home, I don’t have one unified period of getting things done. I have several mini periods. Work isn’t a contiguous landmass of focus; it’s more like an archipelago of productivity amid a sea of…
如何亿点点降低语音识别跨领域、跨语种迁移难度？
作者：侯汶昕、王晋东 编者按：随着深度学习的不断发展，语音识别技术得到了极大的提升，同时为人们的日常生活提供了许多便利。然而，一个语音模型的训练并非易事，因为语音数据天然存在着获取难、数据标注耗时昂贵的问题，而且还会面临模型漂移、标注数据不足等难题。因此，迁移学习技术对于语音数据非常重要。为了解决语音识别的跨领域和跨语言问题，微软亚洲研究院机器学习组和微软（亚洲）互联网工程院提出了跨领域和跨语言语...
Jigsaw fixes bugs in machine-written software
| Rahul Sharma, Suresh Parthasarathy, Sriram Rajamani, Nagarajan Natarajan, and Arun Iyer
Large pre-trained language models such as GPT-3, Codex, and others can be tuned to generate code from natural language specifications of programmer intent. Such automated models have the potential to improve productivity for every programmer in the world. But since…
Just Tech: Centering Community-Driven Innovation at the Margins episode 2 with Dr. Tawanna Dillahunt, Zachary Rowe, and Joanna Velazquez
In “Just Tech: Centering Community-Driven Innovation at the Margins,” Senior Principal Researcher Mary L. Gray explores how technology and community intertwine and the role technology can play in supporting community-driven innovation and community-based organizations. Dr. Gray and her team are…