快猫短视频

Many people who use a voice assistant, such as Alexa or Google Home, will be familiar with them not fully grasping commands. But it appears they may聽be worse at understanding women than men.

Polling company YouGov asked 1000 people in the UK about voice assistants. Around two-thirds of the female participants said the devices failed to respond to their voice commands some of the time, compared with half of the men.

鈥淥ur research reveals that women are more likely to encounter problems being understood by a smart speaker than men, which is ironic, given that most smart speakers are designed to have female voices,鈥� says Russell Feldman at YouGov.

However, as the survey was self-reported, it may not fully match reality, says Trevor Cox at the University of Salford. To confirm the results, tests should be run with different voices and different speech recognition systems, he says.

Earlier studies have suggested gender . Rachael Tatman, while at the University of Washington, recently found that YouTube鈥檚 automatic captioning system, which works using the same principles of a voice assistant, was .

Background noise

Tatman later discovered that much of the difference disappeared , suggesting voice recognition technology may work less well for women in noisy situations.

Separate has found that analysing speech from men and women differently could increase the accuracy of voice recognition systems, suggesting that those designed for one gender won鈥檛 perform as well for the other. Similar accuracy gaps are seen .

Part of the difference in accuracy could be that men speak more loudly. As a result, women鈥檚 voices are more likely to be masked by environmental sounds that make them more difficult to understand, says Tatman

A gender gap could also be caused by limited examples of female voices in the data sets that systems are trained on, a common problem with AI that can introduce bias, says Muneera Bano at the Swinburne University of Technology, Australia. 鈥淚f the data had less examples of female voice, or other dialects and accents, the system wouldn鈥檛 know how to process it.鈥�

It isn鈥檛 possible to know the range of voices included in the databases used by Amazon, Google and Apple, as they aren鈥檛 public. However, Mozilla鈥檚 Common Voice project aims to gather as many different voices as possible, with the collection open to anyone to use.