semantic-pdf-search 0.8.1

Search PDFs semantically

semantic-pdf-search

A semantic PDF searching application, written in Python.

By Jordan Zedeck and Jonathan Louis

Overview

This application utilizes a machine learning embedding model to encode both a PDF document and a user's queries. This process enables the application to find near-matches to the query within the document, much like an internet search engine would for web-pages. The page number results are displayed as buttons which can be clicked to open the PDF directly to the page in your default web browser.

Features

• Semantic Search: Finds near-matches and related concepts, not just exact keywords.
• Offline Operability: Once semantic-pdf-search is installed and used once, it can be used completely offline.
• Cross-Platform: Supports Linux, Windows and macOS.

Example

Query:

Result:

Installation

Prerequisites

This package requires Tkinter. If you run the command python -m tkinter and a new window does not appear, you will need to install it manually.

• Windows: Re-run the Python installer and ensure the tcl/tk checkbox is ticked.
• macOS: Install Tkinter using Homebrew with the following command:

  brew install python-tk
  

• Linux: Varies depending on package manager:
  • Debian:

    sudo apt install python3-tk
    

  • Fedora:

    sudo dnf install python3-tkinter
    

  • Arch: (note that pip installing packages on Arch requires using a venv):

    sudo pacman -S tk
    

From PyPI

The easiest way to install the package is using pip.

pip install semantic-pdf-search

From Source

To install from the GitHub repository, follow these steps:

cd semantic-pdf-search
python -m build
pip install dist/semantic_pdf_search-0.8.0-py3-none-any.whl

Launching semantic-pdf-search

Once installed, run the application from your command line:

semantic-pdf-search

Basic Usage Guide

1. Browse for a PDF: Click on "File ..." -> "Open ..." -> "Browse for PDF" to browse for a PDF file.
2. Select and Open: Navigate to your PDF, select it, and click "Open".
3. Wait for Embeddings: The application will process the document and create embeddings. This may take a moment, especially for large files.
4. Enter a Query: Once the document is loaded, type your query into the search bar and press Enter.
5. View Results: The application will display a list of page numbers that contain near-matches to your query.
6. Open the Page: Click on any of the result buttons to open the PDF directly to that page in your default web browser.