EDS-PDF provides a modular framework to extract text information from PDF documents.

You can use it out-of-the-box, or extend it to fit your specific use case. We provide a pipeline system and various utilities for visualizing and processing PDFs, as well as multiple components to build complex models:complex models:

• 📄 Extractors to parse PDFs (based on pdfminer, mupdf or poppler)
• 🎯 Classifiers to perform text box classification, in order to segment PDFs
• 🧩 Aggregators to produce an aggregated output from the detected text boxes
• 🧠 Trainable layers to incorporate machine learning in your pipeline (e.g., embedding building blocks or a trainable classifier)

Visit the 📖 documentation for more information!

Install the library with pip:

pip install edspdf

Let's build a simple PDF extractor that uses a rule-based classifier. There are two ways to do this, either by using the configuration system or by using the pipeline API.

Create a configuration file:

[pipeline]
pipeline = ["extractor", "classifier", "aggregator"]

[components.extractor]
@factory = "pdfminer-extractor"

[components.classifier]
@factory = "mask-classifier"
x0 = 0.2
x1 = 0.9
y0 = 0.3
y1 = 0.6
threshold = 0.1

[components.aggregator]
@factory = "simple-aggregator"

and load it from Python:

import edspdf
from pathlib import Path

model = edspdf.load("config.cfg")  # (1)

Or create a pipeline directly from Python:

from edspdf import Pipeline

model = Pipeline()
model.add_pipe("pdfminer-extractor")
model.add_pipe(
    "mask-classifier",
    config=dict(
        x0=0.2,
        x1=0.9,
        y0=0.3,
        y1=0.6,
        threshold=0.1,
    ),
)
model.add_pipe("simple-aggregator")

This pipeline can then be applied (for instance with this PDF):

# Get a PDF
pdf = Path("/Users/perceval/Development/edspdf/tests/resources/letter.pdf").read_bytes()
pdf = model(pdf)

body = pdf.aggregated_texts["body"]

text, style = body.text, body.properties

See the rule-based recipe for a step-by-step explanation of what is happening.

If you use EDS-PDF, please cite us as below.

@software{edspdf,
  author  = {Dura, Basile and Wajsburt, Perceval and Calliger, Alice and Gérardin, Christel and Bey, Romain},
  doi     = {10.5281/zenodo.6902977},
  license = {BSD-3-Clause},
  title   = {{EDS-PDF: Smart text extraction from PDF documents}},
  url     = {https://github.com/aphp/edspdf}
}

We would like to thank Assistance Publique – Hôpitaux de Paris and AP-HP Foundation for funding this project.