Peer-reviewed research

Deaf, Hard of Hearing, and Hearing Perspectives on using Automatic Speech Recognition in Conversation

Abraham Glasser et al., ASSETS 2017 (also arXiv 1909.01176), 2017

Read on October 16, 2026

What it is: A study from the Rochester Institute of Technology comparing deaf, hard of hearing and hearing people's views of automatic speech recognition captions in group conversation. Its introduction reviews how much information deaf people get from real-time text.
How much weight it carries: Peer-reviewed conference paper from a university research group. The information-received figures are cited from earlier research in the introduction, not measured by these authors.

Facts taken from this source

The authors note that automatic speech recognition accuracy depends on the speech itself, ambient noise and microphone quality, as well as on computing power.
“ASR accuracy for any application can be affected by other variables including auditory factors such as speech fidelity, ambient noise and microphone quality” (Section 4, ASR for Conversational Use)
The authors identify text accuracy and lag time as the major obstacles to using speech recognition in group conversation for deaf and hard of hearing people.
“Major obstacles that limit ASR use in group conversation for deaf and hard of hearing includes text accuracy and lag time.” (Section 4, ASR for Conversational Use)
The authors report that even with accurate real-time text from human captioners, deaf and hard of hearing people receive 50 to 80 percent of the information, against 84 to 95 percent for hearing peers.
“even when provided with accurate real-time text through captioners, they receive only 50%-80% of the information, compared to 84%-95% for their hearing peers” (Introduction)

Peer-reviewed research

Abraham Glasser et al., ASSETS 2017 (also arXiv 1909.01176), 2017

Read on October 16, 2026

What it is: A study from the Rochester Institute of Technology comparing deaf, hard of hearing and hearing people's views of automatic speech recognition captions in group conversation. Its introduction reviews how much information deaf people get from real-time text.
How much weight it carries: Peer-reviewed conference paper from a university research group. The information-received figures are cited from earlier research in the introduction, not measured by these authors.

The authors note that automatic speech recognition accuracy depends on the speech itself, ambient noise and microphone quality, as well as on computing power.
“ASR accuracy for any application can be affected by other variables including auditory factors such as speech fidelity, ambient noise and microphone quality” (Section 4, ASR for Conversational Use)
The authors identify text accuracy and lag time as the major obstacles to using speech recognition in group conversation for deaf and hard of hearing people.
“Major obstacles that limit ASR use in group conversation for deaf and hard of hearing includes text accuracy and lag time.” (Section 4, ASR for Conversational Use)
The authors report that even with accurate real-time text from human captioners, deaf and hard of hearing people receive 50 to 80 percent of the information, against 84 to 95 percent for hearing peers.
“even when provided with accurate real-time text through captioners, they receive only 50%-80% of the information, compared to 84%-95% for their hearing peers” (Introduction)