developersuryansh/pytorch-chatbot

This is small chatbot can be used in a Cafe, Restaurant, or small business.

🧠 Model

The architecture used was a feedforward network (very simple). The network receives a input vector (length equal to size of bag of words). Subsequently, it shows the user a random response (from the json file) based on that category.

Part of the network code is shown below.

class Net(nn.Module):
    def __init__(self, input_size, hidden_size, output_size):
        super(Net, self).__init__()
        self.linear1 = nn.Linear(input_size, hidden_size)
        self.linear2 = nn.Linear(hidden_size, hidden_size)
        self.linear3 = nn.Linear(hidden_size, output_size)
        self.relu = nn.ReLU()

    def forward(self, x):
        x = self.relu(self.linear1(x))
        x = self.relu(self.linear2(x))
        output = self.linear3(x)
        return output

Architecture of the neural network

For train the network, run:

python train.py

Customize ⚙️

The training dataset used is in the intents.json file, which contains texts in Spanish that will serve as quick responses from our chatbot (A tokenizer and steemer focused on the Spanish language must be used). If you want the chatbot to have more elaborate responses, you can expand the categories and quick responses. Just define a new tag, possible patterns, and possible responses for the chat bot. You have to re-run the training whenever this file is modified.

▶ Demo

If you want to run application, simply install all the dependencies of the requirements.txt and execute the following line.

python app.py

One windows will open immediately, as shown below.