Recognition resources

In ASRaaS, recognition resources are objects that facilitate or improve recognition of user speech. Resources include data packs, domain language models, wordsets, builtins, and speaker profiles.

At least one data pack is required. Other resources are optional.

Data packs

Recognition requests require a data pack.

Builtins

Builtins are slices of the data pack focused on common tasks.

Domain LMs

Domain language models customize the data pack.

Wordsets

Wordsets add terms to the recognition vocabulary.

Speaker profiles

Speaker profiles customize the acoustic model in the data pack.

Resource weights

Recognition uses a weighted mix of resources.

URN format of resources

Resources are identified with URNs.