Spelling Correction Using LLM and Knowledge Retrieval

This script utilizes a combination of large language models (LLM), rule-based spell checkers, and knowledge-based retrieval to correct spelling mistakes in sentences. The correction process involves both a grammar tool (language_tool_python) and a retrieval-augmented generation (RAG) approach using a pre-trained model from the Hugging Face transformers library.

Features

• Corrects common spelling mistakes using a pre-trained language model (GPT-Neo).
• Utilizes a knowledge base to retrieve similar spelling corrections and provide context for the model.
• Includes an initial pass for spelling correction using the language_tool_python library.
• Output is filtered to ensure the correction maintains the original meaning of the text.

Requirements

Before you can run this project, ensure that the following Python packages are installed:

• language_tool_python
• transformers
• scikit-learn
• torch