nlpchat 1.2

A package to simplify chatbot creation using NLP.

Here’s the updated README code with a hybrid approach using both Logistic Regression and Neural Networks:

---

nlpchat

nlpchat is a Python package that simplifies the creation of chatbots using natural language processing (NLP) for intent identification. The package uses Sentence Transformers for embedding input text and supports easy management of intents with customizable responses. It now supports a hybrid approach using both Logistic Regression and Neural Networks for enhanced performance in intent classification. It provides functionality for training, saving, and loading models, allowing you to avoid retraining the chatbot each time.

Features

• Simple intent management: Easily add intents, patterns, and responses.
• Hybrid Intent Recognition: Uses both Logistic Regression and Neural Networks for better intent classification.
• NLP-powered: Uses Sentence Transformers to embed and understand user inputs.
• Train and Save: Train the model once and save it for later use.
• Load saved models: Quickly load previously trained models for immediate use.
• Intent prediction: Predicts the user's intent based on input text.
• Custom responses: Generate responses based on detected intents.
• Intent-only prediction: Retrieve just the predicted intent without generating a response.

Installation

You can install the package from the repository:

pip install nlpchat

Usage

Here’s a simple example showing how to create a chatbot, train it using a hybrid approach, save the model, and make predictions.

1. Import and Initialize the Chatbot

from nlpchat import NlpChat

# Create an instance of NlpChat
chatbot = NlpChat()

2. Add Intents

You can add intents using the add_intent() function by providing a tag (intent name), patterns (user inputs), and responses.

# Add a greeting intent
chatbot.add_intent(
    tag="greeting",
    patterns=["Hi", "Hello", "Good morning", "How are you?"],
    responses=["Hello!", "Hi there!", "Good to see you!"]
)

# Add a goodbye intent
chatbot.add_intent(
    tag="goodbye",
    patterns=["Bye", "See you later", "Goodbye"],
    responses=["Goodbye!", "See you later!", "Take care!"]
)

3. Train the Model

Train the chatbot on the added intents. The hybrid model uses both Logistic Regression and Neural Networks for better intent prediction.

chatbot.train()

4. Save the Model and Intents

Once the model is trained, you can save it to a file for future use.

# Save the trained model and intents to a file
chatbot.save_model("chatbot_model.pkl")

5. Load a Saved Model and Intents

To avoid retraining every time, load the saved model.

# Load the saved model
chatbot.load_model("chatbot_model.pkl")

6. Predict Intent and Get Response

After the model is loaded, you can use it to predict user intents and get a response.

user_input = "Hello"
response = chatbot.get_response(user_input)
print(response)  # Outputs: "Hello!" or another greeting response

7. Predict Intent Only

If you only need to predict the user's intent without generating a response, you can use the get_intent() method:

# Get only the predicted intent (without response)
user_input = "Hello"
intent = chatbot.get_intent(user_input)
print(f"Predicted intent: {intent}")  # Outputs: "greeting"

How It Works

• Intent Management: Users define intents using add_intent(). Each intent has a tag (such as "greeting"), a set of patterns (user inputs), and a set of responses.
• Hybrid Approach: The chatbot combines Logistic Regression and Neural Networks. Sentence Transformer embeddings are passed through Logistic Regression for initial classification and then fine-tuned by the Neural Network for intent prediction.
• Prediction: When user input is given, the chatbot encodes the input using the Sentence Transformer model, predicts the intent using the hybrid model, and returns a random response from the associated intent (or just the intent if requested).

Dependencies

The following dependencies are required:

• sentence-transformers
• scikit-learn
• tensorflow
• numpy
• pickle-mixin

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contributing

Contributions are welcome! If you’d like to contribute to the project, please fork the repository and submit a pull request.

Contact

For any issues or suggestions, please open an issue on the GitHub repository.

---

This version reflects the hybrid approach for better performance using both Logistic Regression and Neural Networks. Let me know if any further adjustments are needed!