Virtual Assistant

Features

• Speech Recognition: Utilizes the Vosk model for accurate speech recognition, enabling seamless interaction through voice commands.
• Intent Classification: Employs a three-layer linear model with ReLU activation functions to classify user intents effectively.
• Basic Task Automation: Capable of performing various basic tasks, such as creating, renaming, and deleting folders, searching the web, managing device controls, and more.
• Expandability: Can be easily expanded to accommodate additional functionalities based on specific user needs.

Usage

To use the Virtual Assistant:

• Download Speech Recongition Model: Extract the compressed .rar file inside model named directory.

  Click here

• Install Dependencies:

  pip install -r requirements.txt

• Run train.py for Intent Classification:

  py .\neuralnet\train.py -e <EPOCHS> -b <BATCH_SIZE> -p <PLOT:1>  

  Argument	Short Option	Type	Default	Help
  --batch_size	-b	int	8	Batch size for training
  --epochs	-e	int	250	Number of epochs for training
  --plot	-p	int	0	Whether to plot the loss curve (1 for yes, 0 for no)

  The intent.pth will be saved under model directory after executing train.py.

• Set API Keys:

  Create a .env file in the main directory of the project. Inside the .env file, define the variable for your API key

  Below is a list of websites and APIs used in this project. Click on the links to access their documentation and obtain the necessary information.

  • Open Weather Map

    Make sure to review the documentation for each API to understand their usage and any specific requirements, such as obtaining API keys or authentication tokens.

  WEATHER_API_KEY = "{{secret.YOUR_API_KEY}}"

• Run Script:

  py chat.py

• Interaction:

  You: hello
  ByteBot: Hi there, how can I assist you today?
  You: tell me a joke
  ByteBot: A perfectionist walked into a bar...apparently, the bar wasn't set high enough
  You:  
  

• Customization:

  If you want to customize the intent according to your needs, add the tags, patterns and responses inside intents.json in following format and re-run train.py while adding functionality features in chat.py. The intent.json should follow this base format:

  {
      "intents": [
          {
              "tag": "greeting",
              "patterns": [
                  "Hi there",
                  "Hello",
                  "yo"
              ],
              "responses": [
                  "Hello",
                  "Good to see you again",
                  "Hi there, how can I help?"
              ],
          }
      ]
  }

Speech Recognition Model

If the speech recognition model shows errors, run this script on the same directory path

from engine import initialize_model, speech_recognize
# Transcript
transcript = ''

if __name__ == "__main__":
    recognizer, mic, stream = initialize_model()

    while True:
        data=stream.read(4096)
        if recognizer.AcceptWaveform(data):
            result = recognizer.Result()[14:-3]
            transcript += result + ' '
            print(transcript)

Contributors

Limitations

The current version of the Virtual Assistant may have limitations in complex tasks or specialized domains. You can add more functions or integrate Language Models like Llama2, MistralAI, BERT.

---

Feel free to customize and extend the Virtual Assistant to suit your specific needs and requirements. Contributions and feedback are welcome!