invoice-field-extractor

Extract some known fields of invoices such of organization name, date, invoice amount, using machine learning

Dependencies

You should have these dependencies to run this project. I only demonstrate installation for Ubuntu, tested on Ubuntu 16.04. For other operating system, installation will vary.

Tesseract

Install Tesseract via APT

sudo apt update sudo apt install tesseract-ocr

Then, clone tessdata project from Tesseract Github repo. Usually we change directory to /usr/share/tesseract-ocrfirst, then clone the project there

git clone https://github.com/tesseract-ocr/tessdata

Add the parent directory of this cloned project to $TESSDATA_PREFIX environment variable. Add the $TESSDATA_PREFIX with parent directory value to /etc/environment so we can get rid of this env variable later.

Stanford NER Tagger

Download Stanford NER Tagger from this link.

https://nlp.stanford.edu/software/CRF-NER.shtml#Download

And then unzip. Add this directory after you unzip to $PATH environment variable. You can add this path to /etc/environment. Please make sure that you have permission to execute ner.sh script, otherwise please chmod first. And make sure ner.sh can be called from any directories.

PDFLib TET

Download from this link

http://www.pdflib.com/download/tet/

Unzip the file. Then add <directory you unzip the file>/bin to $PATH environment variable. Make sure that binary file tet can be executed from any directories.

Usage

Open this project in Intellij IDEA. Add JAR dependencies from /lib to the project. And then build artifact.

You can run the JAR by:

java -jar invotract.jar