FileStruct

FileStruct is a high-level Python library that aims to extract the overall structure of documents, particularly PDFs, based on visual information such as size, color and font.

How does it work ?

As clever human beings, we are able to detect titles, subtitles, and paragraphs using the visual appearence of the document. A big text in red most certainly represent a title (or subtitle). Using these heuristics, we are able to structure a document : This paragraph belongs to this section. The same method is used by this package to provide an automated, while realistic way to structure a document. The method is described bellow :

1. Text and style extraction : We rely on lower level librairies (like PyMuPDF) for the extraction of the text and style information, and the ordering of each block of text.
2. Tree creation : A tree is created, in which each block of text is a node of the tree. A child of a node in the tree is a subsection of a section in the document.
3. Data exportation : The data can be exported in JSON format.

For now, filestruct can only read formats that are supported by PyMuPDF. This includes pdf, epub, xps, mobi, fb2, cbz and svg. I plan to add more file formats in the future.

Installation

Install FileStruct using pip :

pip install filestruct

Getting Started

Bellow, a basic usage for a PDF document :

from filestruct.document import PDFDocument

doc = Document("PATH_TO_YOUR_FILE.pdf")
data = doc.to_json()   # Export the tree into json format
print(data)
print(doc)