Zen Document Parser

Intro

zen_document_parser is a utility for extracting data from various official documents. It uses PDFQuery behind the scenes.

Currently, there is out-of-the-box support for parsing Indian Government ITR-V PDF documents.

The library also supports parsing of arbitrary PDF documents by allowing you to specify a 'schema' for the document. The library allows for multiple 'variants' of a document. For example, The Indian ITR-V document has slightly different fields and layout depending on whether it was generated in 2013, 2014, 2015 etc.

Check out the examples below.

Installation

Install using pip like so:

$ pip install zen_document_parser

Usage

ITR-V Docs

from zen_document_parser.itr.itr import ITRVDocument

# You can pass in a path or a file-like object during instantiation.
doc = ITRVDocument('/path/to/itrv.pdf')

# Will load the file, auto-detect the variant and perform extraction of all
# fields and store results internally.
doc.extract()

# Extracted fields are available in the `data` property.
print(doc.data.company_name)
print(doc.data.gross_total_income)

Configuring for custom PDF documents

You basically follow these steps:

• Define one or more 'schemas', ie. DocVariant subclasses, to go with each variant of the doc.
• In each of these variants, define a check_for_match() method that returns True if a file was successfully parsed.
  • Make sure to define test_fields as an attribute on each class that is a list of all field names used inside check_for_match(). (This is required at present for optimization purposes, but will not be a requirement in an upcoming version.)
• Define a Doc subclass that represents your document. In the variants attribute, specify possible variants.

from zen_document_parser.base import DocField, DocVariant, Document


class Variant1(DocVariant):

    # The fields that are used inside `check_for_match()`. (for optimization)
    test_fields = ['form_title']

    form_title = DocField((30, 300, 500, 380))
    name = DocField((100, 120, 400, 140.5))
    address = DocField((150, 90, 650, 110))

    def check_for_match(self):
        if self.form_title == 'Application Form For 2014':
            return True
        return False


class Variant2(DocVariant):

    test_fields = ['form_title']

    form_title = DocField((30, 290, 500, 380))
    name = DocField((70, 140, 350, 160))
    address = DocField((150, 120, 650, 140))
    pan_no = DocField((150, 80, 650, 100))

    def check_for_match(self):
        if self.form_title == 'Application Form For 2015-16':
            return True
        return False


class MyForm(Document):

    variants = [Variant1, Variant2]


def main():
    doc = MyForm('/path/to/form.pdf')
    doc.extract()
    print(doc.data.to_dict())

TODO

• Hanle data-type specification
• Handle fields being mandatory/non-mandatory.
• Right now the user has to explicitly specify test_fields for optimization purposes. Find a way where this isn't needed.
  • Automatically load them the first time they're referred to? extract() can still be there as a way to bulk-load all fields in one go.