Hotword recognition in speeches - Intelligent Media Services

The built-in speech-to-text (STT) model allows you to configure hotwords. This topic describes the hotword feature and how to configure hotwords.

Feature description

The hotword feature allows you to specify a set of words as hotwords. When the STT model is running, the model processes these hotwords first and increases their weight in the recognition process. If you find that the recognition results of specific words of your business do not meet your requirements, you can use the hotword feature to improve the recognition results.

Configuration process

Prepare a hotword file: Write the keywords that you want to recognize into a TXT file. Example:
```
The Shawshank Redemption
Farewell My Concubine
Léon: The Professional
Forrest Gump
Life Is Beautiful
Titanic
Spirited Away
Schindler's List
Inception
WALL-E
```
Note
- The uploaded hotword file can be up to 100 KB in size. The file can contain up to 500 words, and each word can be up to 10 characters in length.
- Business-specific hotwords must be encoded in UTF-8 and cannot contain punctuation marks or special characters.
Upload a hotword file to the STT node: Upload your hotword file to the STT node of the real-time workflow.
The hotword file takes effect: The uploaded hotword file takes effect only for the workflow that is saved or created after the upload.