LSTM seq2seq chatbot

LSTM Seq2Seq Chatbot

This project implements a chatbot using a Sequence-to-Sequence (Seq2Seq) architecture with Long Short-Term Memory (LSTM) networks. The model is trained on a custom dataset of domain-specific input-output conversations (e.g., furniture-related queries) to generate human-like responses. It utilizes an encoder-decoder architecture where the encoder processes the input sentence and the decoder generates the corresponding response. The model also incorporates techniques such as tokenization, padding, and teacher forcing to improve performance. The chatbot is capable of understanding and responding to user inputs in natural language, making it suitable for integration into customer support systems or virtual assistants.

Key Features:

Encoder-Decoder architecture with LSTM layers

Byte Pair Encoding (BPE) for subword tokenization

Custom domain-specific training data

Support for saving and loading the trained model and tokenizer

Designed for deployment with FastAPI or similar frameworks