Rhino Speech-to-Intent Engine

Made in Vancouver, Canada by Picovoice

Rhino is Picovoice's Speech-to-Intent engine. It directly infers intent from spoken commands within a given context of interest, in real-time. For example, given a spoken command:

Can I have a small double-shot espresso?

Rhino infers that the user would like to order a drink and emits the following inference result:

{
  "isUnderstood": "true",
  "intent": "orderBeverage",
  "slots": {
    "beverage": "espresso",
    "size": "small",
    "numberOfShots": "2"
  }
}

Rhino is:

• using deep neural networks trained in real-world environments.
• compact and computationally-efficient, making it perfect for IoT.
• self-service. Developers and designers can train custom models using Picovoice Console.

Compatibility

• Python 3.9+
• Runs on Linux (x86_64), macOS (x86_64, arm64), Windows (x86_64, arm64), and Raspberry Pi (Zero, 3, 4, 5).

Installation

pip3 install pvrhino

AccessKey

Rhino requires a valid Picovoice AccessKey at initialization. AccessKey acts as your credentials when using Rhino SDKs. You can get your AccessKey for free. Make sure to keep your AccessKey secret. Signup or Login to Picovoice Console to get your AccessKey.

Usage

Create an instance of the engine:

import pvrhino

access_key = "${ACCESS_KEY}" # AccessKey obtained from Picovoice Console (https://console.picovoice.ai/)

handle = pvrhino.create(access_key=access_key, context_path='/absolute/path/to/context')

Where context_path is the absolute path to Speech-to-Intent context created either using Picovoice Console or one of the default contexts available on Rhino's GitHub repository.

The sensitivity of the engine can be tuned using the sensitivity parameter. It is a floating-point number within [0, 1]. A higher sensitivity value results in fewer misses at the cost of (potentially) increasing the erroneous inference rate.

import pvrhino

access_key = "${ACCESS_KEY}" # AccessKey obtained from Picovoice Console (https://console.picovoice.ai/)

handle = pvrhino.create(access_key=access_key, context_path='/absolute/path/to/context', sensitivity=0.25)

When initialized, the valid sample rate is given by handle.sample_rate. Expected frame length (number of audio samples in an input array) is handle.frame_length. The engine accepts 16-bit linearly-encoded PCM and operates on single-channel audio.

def get_next_audio_frame():
    pass

while True:
    is_finalized = rhino.process(get_next_audio_frame())

    if is_finalized:
        inference = rhino.get_inference()
        if not inference.is_understood:
            # add code to handle unsupported commands
            pass
        else:
            intent = inference.intent
            slots = inference.slots
            # add code to take action based on inferred intent and slot values

When done resources have to be released explicitly:

handle.delete()

Non-English Contexts

In order to run inference on non-English contexts you need to use the corresponding model file. The model files for all supported languages are available here.

Train Models over API

You can train models over API without going to the console:

train_context_from_dynamic_slots(
        "${ACCESS_KEY}",                             # AccessKey obtained from Picovoice Console (https://console.picovoice.ai/)
        "${OUTPUT_PATH}",                            # Path to save the newly trained model
        "${LANGUAGE}",                               # Two-character language code
        "${CONTEXT_PATH}",                           # Path to Rhino's context (.rhn) file
        {"${SLOT_NAME}": {"${SLOT1}", "${SLOT2}"}},  # Dynamic slot key-value pairs to merge into the YAML's `context.slots` section.
        "${PLATFORM}"                                # Optional platform for the trained model. If None, the default(current) platform is used.

(or)

train_context_from_yaml(
        "${ACCESS_KEY}",                             # AccessKey obtained from Picovoice Console (https://console.picovoice.ai/)
        "${OUTPUT_PATH}",                            # Path to save the newly trained model
        "${LANGUAGE}",                               # Two-character language code
        "${YAML_PATH}",                              # Path to YAML configuration file
        "${PLATFORM}"                                # Optional platform for the trained model. If None, the default(current) platform is used.

train_context_from_dynamic_slots is better suited if you would like the add additional slot values to your current Rhino model.

Check Rhino Model API docs for a list of supported languages and platforms.

Demos

pvrhinodemo provides command-line utilities for processing real-time audio (i.e. microphone) and files using Rhino.