The LLM Semantic Book Recommender demonstrates how to build a semantic search system for book recommendations. The project utilizes pre-trained transformer models to generate vector embeddings from book metadata. By indexing these embeddings, the system efficiently returns recommendations that align with the user's search query—even if the phrasing does not match the dataset keywords directly.
Notebooks/Data_Preprocessing Text data cleaning.This involved removing extraneous punctuation, stop words and normalizing text.
Notebooks/vector_search Semantic (vector) search and how to build a vector database. This allows users to find the most similar books to a natural language query (e.g., "a book about a person seeking revenge"). Text data (book descriptions) is converted to vector embeddings using pre-trained transformer model. This process captures the semantic meaning of the text.
Notebooks/text_classification Doing text classification using zero-shot classification in LLMs. This allows us to classify the books as "fiction" or "non-fiction", creating a facet that users can filter the books on.
Notebooks/Sentiment_Analysis Doing sentiment analysis using LLMs and extracting the emotions from text. This will allow users to sort books by their tone, such as how suspenseful, joyful or sad the books are.
gradio_dashboard Creating a web application using Gradio for users to get book recommendations .
- Clone the Repository:
git clone https://github.com/KimathiNewton/llm_book_recommender.git
cd llm_book_recommenderpython -m venv venv
source venv/bin/activate # On Windows use: venv\Scripts\activateInstall Dependencies:
pip install -r requirements.txtThe repository includes a Gradio-based dashboard to interact with the recommender system. To launch the dashboard, run:
python gradio_dashboard.pyThis command will start a local web server. Open the provided URL in your browser and follow the instructions to input your preferences and see book recommendations.