I'm thinking about how the gender, talking speed and accent of the speech voice used affects user understanding. (UK market) Examples of speech interface: SatNavs, Automated Telephone Assistance Services, etc.
Yes. In regards to research on which gender to use, I suggest you read a book called, "Wired for Speech" by Cliff Nass.
I work as a voice user interface designer and I've compiled some lists:
- Understand your target audience before starting any design work.
- Use the brand’s persona for creating and playing prompts.
- Use conversational language rather than treating the system as “information collector”.
- Try not to use “Please” on every single line.
- Try not to use “Sorry” too much.
- Understand your users’ mental model and cognitive limitations when presenting information. For example, native English speakers often place new information at or near the end of the sentence, and this is where they often expect to retrieve it.
- Use cohesion devices such as pronouns, discourse markers, pointer words, etc.
- Remove words and phrases that can be supplied by context.
There are other conditions to think about such as error strategies.
What I often I see in a lot of automated systems, including virtual assistants is the lack of escalation on dialog error prompts. For example, I was interacting with a bot some time ago and when I said something out of context, it just responded with, “Sorry I’m not smart enough!”
First of all, never acknowledge you’re stupid. It’s OK to acknowledge you don’t know the answer but admitting you’re “not smart enough” is too much; even for a bot :)
Here’s a way for dealing with escalating error prompts:
- System: Thanks for calling. How can I help you?
- Caller: Uhhhhhhh
- System: You can say – get order status, account information, or just say – it’s something else.
- Caller: Get order status
- System: OK – what’s the order number? If you don’t have it, just say – I don’t have it. ….
Simple eh? So not sure why a lot of these bots and systems miss this. A great user experience happens when you take the user back on track by guiding them.
Now you don’t have to give all the details right away. There’s another approach known as rapid re-prompt. This approach doesn’t give away all the details but instead, the response is given with something simple, such as, “I’m sorry?”. Here’s an example:
- System: Tell me your date of birth.
- Caller: Well…
- System: I’m sorry?
The disadvantage as you know is that the user is not getting all the information in details. The best place to use this is during open-ended prompts, such as ‘How can I help you?’ and the user is often not considering this as an error.
Somewhat old paper, but has a decent starting point: User Interface Design Guidelines for Speech
When and how to develop a speech recognition application:
- Use speech when the user’s eyes and hands are busy or when the user is moving.
- Train the speech recognition system in the user’s work environment.
- Iteratively evaluate and re-design the speech recognition application.
10 guidelines related to the software portion of the speech recognition user interface:
- Keep small the number of words in the speech recognition vocabulary.
- Keep short each speech input.
- Use speech inputs that sound distinctly different from each other.
- Provide immediate feedback for every speech input.
- Keep the user interface simple.
- Make error correction intuitive.
- Avoid modes.
- Don’t use speech to position objects.
- Use a command-based user interface.
- Allow users to quickly and easily turn off and on the speech recognizer.
However, keep in mind that natural language processing has come a long way in the past couple years making it possible to have seamless voice interaction rather than just command based voice input.