Recognizers

A recognizer is short for Automatic Voice Recognizer (ASR) and is a service used to parse audio files for text. So, the input of recognizer is audio files and the output is identified text.

Changing recognizers

On the Furhat web interface, on the left hand side, you can adjust which Recognizer Furhat is using. By default, Furhat uses the on-board (i.e offline) windows recognizer.

You are recommended to sign up for either Google or NuanceCloud recognition services since these offer a much better recognizition. Once you have registered for these services, you can put your credentials into this page, save it, then use either one these other recognizers with your flow.

Note: make sure your skill is set up to use an open recognizer when you use Google and Nuance recognizers. Instructions how to do this here Configuring skill

Supported recongizers

Furhat currently comes with support for several recognizers out of the box. There are two types of recognizers, open and grammar. When you use a grammar based, the speech needs to match exactly what the user said. When you use an open recognizer, you can listen to any phrase and utterance. The open recognizers supported are Google and Nuance, both of them are cloud-based, which means that you have to create an account, and have an Internet connection when the recognizer is running.

Google - open recognizer

Google's cloud recognizer is an open recognizer and can hence detect any word in the chosen language. If you want to use Google, you need to sign up for a developer account here. You then need to create a Service Account and download the JSON key file. Open the file in a text editor and paste the json-snippet into the textbox on the recognizer page in the web interface.

Note, when you download your credentials, make sure you select a service account and format json.

Nuance - open recognizer

Nuance's cloud recognizer is an open recognizer and can hence detect any word in the chosen language. To sign up for it, follow the instructions here to get your APP_ID and APP_KEY.

Windows - grammar based recognizer

The on-board and default recognizer. This recongizer is grammar-based, i.e you need to supply the grammar (words and phrases) that it should listen to. In order to use the Windows recognizer you need to have the correct Windows Language Pack for the specific language. It is recommended to not use this recognizer unless you have to - for example if there is no available internet connection.