Describe the book you want, and the mood. It finds it by meaning.

role
Solo build
status
Shipped
when
2025
ownership
Solo
Repository
books.vecAI research · Data & ML engineering
“a slow, hopeful story about a family”joycalmsuspensesadsurprise

Who did what. Personal project.

Overview

Books are embedded by their descriptions and searched by meaning; zero-shot classification adds categories and sentiment analysis tags each book's emotional tone, so results can be re-ranked toward happy, sad, suspenseful or surprising.

Why it exists

Metadata search can't find 'a slow, hopeful story about a family'. Meaning and tone can.

My part

  • Data cleaning, vector search, zero-shot categories, emotion tagging and a Gradio dashboard.

How it works

  1. a

    Clean

    Kaggle books dataset prepared.

  2. b

    Embed

    Hugging Face embeddings and vector search.

  3. c

    Tag

    Zero-shot categories and emotion scores.

  4. d

    Rank

    Query, category and tone.

Built with

  • Python
  • Hugging Face
  • Vector search
  • Gradio

Status & next

Runs as a Gradio dashboard.

Source: github.com/AA-KH/Semantic-Book-Recommender