The speech recognition services may fail to recognize some unique words in your business field. In this case, you can use the hotword feature.
A 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, hotwords can neither contain too many characters in length nor contain punctuation marks. There are no strict restrictions on the length of a hotword. After a hotword is set, you can test the speech recognition effect, and make adjustments if unsatisfactory.
Timeliness of hotwords: After a hotword is set, it immediately takes effect for new recognition requests. However, it cannot be used by ongoing recognition requests.
Hotwords can be divided into extended hotwords and categorized hotwords.
Extended hotwords:In the console, the hotwords in the Others category for a project are extended hotwords. You can add words that are poorly recognized to a hotword vocabulary to improve the recognition effect of these words. There are no category restrictions on extended hotwords. For example, you can addapple, husky, and Xiaoming to the same hotword vocabulary.
Categorized hotwords:Similar to extended hotwords, categorized hotwords are also added to hotword vocabularies to improve the recognition effect. Categorized hotwords are divided into the Name and Place categories. A categorized hotword vocabulary provides a better recognition effect than an extended hotword vocabulary.
- The hotword file must be in the TXT format. It must be encoded in UTF-8 without the byte order mark (BOM). The size of the file cannot exceed 100 KB.
- One hotword occupies one line. A maximum of 128 hotwords can be added. Each hotword must be 1 to 10 characters in length.
- You can add a maximum of 10 categorized hotword vocabularies. That is, you can add a maximum of 10 hotword vocabularies for the Name and Place categories in total. You can add a maximum of 10 extended hotword vocabularies. That is, you can add a maximum of 10 hotword vocabularies for the Others category.
- You need to spell out the numbers in hotwords. For example, you need to convert 58.9 to fifty-eight point nine.
- No special characters other than the space character, tab, line break, and page break are allowed in the file content.
- Categorized hotwords are applicable only to the models that use the administration and court model as the basic model.
- Currently, only Chinese hotwords can be recognized.
In the Create Hotword dialog box, set relevant parameters, upload a hotword file, and then click OK. If no problems are found in the hotword file, a hotword vocabulary is generated. In this example, a hotword file of the Name category is uploaded. Select the hotword vocabulary from the Name drop-down list, and then click Publish for the keyword vocabulary to take effect. Ongoing recognition tasks cannot use this hotword vocabulary.
Note: The hotwords specified for a project in the console are bound to the appkey of the project. The server automatically determines and uses these hotwords based on the appkey on the client without the need of manual settings.
To manage existing hotword vocabularies, choose My Resources > Hotwords in the left-side navigation pane. Take the Name category as an example. The value of Associated Project for a hotword vocabulary indicates the projects that are using this hotword vocabulary. One hotword vocabulary can be used by multiple projects. You can update a hotword vocabulary by re-uploading a hotword file, download a hotword file, or delete a hotword vocabulary.
You can also click Create Hotword to add hotwords. The procedure is the same as that for adding hotwords during project configuration.