Preprint

Achieving Human Parity in Conversational Speech Recognition

W. Xiong et al., arXiv (Microsoft Research), 2017

Read on October 16, 2026

What it is: Microsoft Research measured the word error rate of professional transcribers on the NIST 2000 conversational telephone test set and reported that its automated system matched or slightly beat them. It is the best known milestone for machine accuracy on conversational speech.
How much weight it carries: Widely cited and later peer-reviewed. It measures one benchmark of telephone conversations in English; the human baseline came from one transcription service, and later work found humans can do better with more effort.

Facts taken from this source

Microsoft Research reported in 2016 that its system reached human parity, and professional transcribers had a 5.9 percent error rate on the Switchboard portion.
“find that our latest automated system has reached human parity. The error rate of professional transcribers is 5.9% for the Switchboard portion of the data” (Abstract)
On the NIST 2000 Switchboard test of telephone conversations, professional human transcribers had a word error rate of 5.9 percent, and 11.3 percent on the harder CallHome conversations.
“The error rate of professional transcribers is 5.9% for the Switchboard portion of the data, in which newly acquainted pairs of people discuss an assigned topic, and 11.3% for the CallHome portion” (Abstract)
Microsoft's automated system reached word error rates of 5.8 percent on Switchboard and 11.0 percent on CallHome, slightly below the professional transcribers.
“our automated system establishes a new state of the art, and edges past the human benchmark, achieving error rates of 5.8% and 11.0%, respectively” (Abstract)

Preprint

W. Xiong et al., arXiv (Microsoft Research), 2017

Read on October 16, 2026

What it is: Microsoft Research measured the word error rate of professional transcribers on the NIST 2000 conversational telephone test set and reported that its automated system matched or slightly beat them. It is the best known milestone for machine accuracy on conversational speech.
How much weight it carries: Widely cited and later peer-reviewed. It measures one benchmark of telephone conversations in English; the human baseline came from one transcription service, and later work found humans can do better with more effort.

Microsoft Research reported in 2016 that its system reached human parity, and professional transcribers had a 5.9 percent error rate on the Switchboard portion.
“find that our latest automated system has reached human parity. The error rate of professional transcribers is 5.9% for the Switchboard portion of the data” (Abstract)
On the NIST 2000 Switchboard test of telephone conversations, professional human transcribers had a word error rate of 5.9 percent, and 11.3 percent on the harder CallHome conversations.
“The error rate of professional transcribers is 5.9% for the Switchboard portion of the data, in which newly acquainted pairs of people discuss an assigned topic, and 11.3% for the CallHome portion” (Abstract)
Microsoft's automated system reached word error rates of 5.8 percent on Switchboard and 11.0 percent on CallHome, slightly below the professional transcribers.
“our automated system establishes a new state of the art, and edges past the human benchmark, achieving error rates of 5.8% and 11.0%, respectively” (Abstract)