As is the case with virtually all Watson services, you can customize your instance of Watson speech. In this case, the Speech to Text service has two models that can be customized—a language model and an acoustic model—both of which will be explained further.
The language model tells Watson about new words to listen out for. In essence, the custom language models expand Watson's vocabulary. You will not provide any acoustic signature for these vocabulary words, but rather you simply create an entry for these words, their display spelling, and its various (one or more) sounds-like phonetic spellings (refer to the Using sounds-like customization section of this chapter).
Watson provides two different ways of adding to its vocabulary. The easier approach is to provide a text document that exemplifies the type of vocabulary that is used in your domain. Watson will parse this document and add any words it does not already...