Note
Please see Azure Cognitive Services for Speech documentation for the latest supported speech solutions.
Phrase Generator Usage Scenarios
Phrase Generator outputs a list of phrases from a grammar based on parameters that you specify.
You can use Phrase Generator to identify whether your grammar generates phrases that are not expected by your voice application (“over-generation”), whether your grammar does not generate phrases that your voice application does expect ("under-generation"), and whether your grammar generates nonsense phrases. In addition, you can generate a random sampling of phrases (based on rule weights) from a grammar to obtain a test set, or simply for ‘sanity checking’ the grammar.
The following are some typical scenarios for using Phrase Generator:
You have authored a grammar and used Grammar Validator to verify that it is syntactically correct. Before you invest in testing with utterances, you want to make sure that your grammar recognizes all you intend, but no more. You need to know the following:
Does my grammar over-generate phrases thereby increasing the risk of a false recognition?
Does my grammar under-generate phrases, increasing the risk of unintended recognition rejections?
Will my grammar recognize nonsense phrases? Can I ‘sanity check’ the phrases to judge whether they are linguistically correct?
Based on the output from another of the grammar tools, you have changed the grammar. Now you need to know:
- How many more phrases are actually being recognized after I make a change?
You want to test your application with actual voice input.
- What phrases should testers use to speak to the application so I can verify that it works correctly?
You want to help the speech recognizer to avoid confusing some phrases with others.
- How can I generate lists of phrases to input to the Confusability tool?