In speech recognition services, if the default recognition result for some business-specific words does not meet your requirements, you can add these words to a vocabulary as extended hotwords to improve the recognition effect.
An extended hotword can be the name of a person or company. It can also be a proper noun in a certain field, such as Tracy (a person name which is usually recognized as Tracey), appellee in the judicial field, and free shipping in the e-commerce field. Generally, extended hotwords can neither contain too many characters in length nor contain punctuation marks. For more information about specific restrictions, see the following sections. After an extended hotword vocabulary is set, you can test the speech recognition effect, and make adjustments if unsatisfactory.
When a universal model is used, only a limited number of extended hotwords can be added. We recommend that you use a custom model if you need to add a large number extended hotwords.
- If you set extended hotwords for a project in the console, the hotwords are bound to the appkey of the project. You do not need to specify the extended hotword vocabulary in the client code. If you create an extended hotword vocabulary by using the corresponding pctowap open platform (POP) API operation, you need to call the corresponding SDK method to specify the ID of the vocabulary in the client code for the vocabulary to take effect.
- If you set extended hotwords for a project in the console, each hotword occupies a line in the training corpus. You do not need to set the weight for each hotword. If you create an extended hotword vocabulary by using the corresponding POP API operation, you need to set the weight for each hotword.
If some movie names often appear in the object to be recognized, you can add them to the hotwords of a project to improve their recognition rate.
The Shawshank Redemption
Farewell My Concubine
Léon: The Professional
Life Is Beautiful
Hachi: A Dog's Tale
The Legend of 1900
The Truman Show
My Neighbor Totoro