When you use speech recognition services, you may find that the default recognition results of words specific to your business field are not as expected. In this case, you can add these words as hotwords to a vocabulary to improve the recognition results.
Hotword classification
Name hotwords
You can add only person names and place names as name hotwords. One hotword vocabulary can contain only person names or place names.
Business-specific hotwords
Business-specific hotwords include the names of persons or companies and proper nouns used in various business fields. Examples are the person name Tracy, which is usually recognized as Tracey, the proper noun "appellee" in the judicial field, and the proper noun "free shipping" in the e-commerce field.
Hotwords cannot contain punctuation marks. Each hotword can contain up to 10 Chinese characters in length.
You can add 10 groups of name hotwords and 10 groups of business-specific hotwords. Each group can contain up to 128 hotwords.
You can associate one group of person name hotwords, one group of place name hotwords, and one group of business-specific hotwords with each project.
To add a large number of hotwords, you can create a custom linguistic model. For more information about custom linguistic models, see Overview.
Comparison between two methods used to create hotwords
If you use the Intelligent Speech Interaction console to create and associate hotwords with the appkey of a project, you do not need to specify a hotword vocabulary in your code. If you use the Alibaba Cloud pctowap open platform (POP) API, you can create vocabularies only of business-specific hotwords. A hotword vocabulary that you create by calling the POP API from a client takes effect only after you call the SDK to set the vocabulary ID in the code.
If you use the Intelligent Speech Interaction console to create hotwords, make sure that each hotword occupies a separate line in the hotword file that you upload. You do not need to specify the weight of each hotword. If you call the POP API to create a vocabulary of business-specific hotwords, you must specify the weight of each hotword.
Sample application
To improve the recognition results of the following movie names, you can add the movie names to a project as hotwords:
The Shawshank Redemption
Farewell My Concubine
Léon: The Professional
Forrest Gump
Life Is Beautiful
Titanic
Spirited Away
Schindler's List
Inception
WALL-E
Hachi: A Dog's Tale
3 Idiots
The Legend of 1900
The Chorus
The Truman Show
The Godfather
My Neighbor Totoro
Interstellar