Skip to content

Latest commit

 

History

History
48 lines (30 loc) · 4.36 KB

index.md

File metadata and controls

48 lines (30 loc) · 4.36 KB
layout title
default
PennyLedger

What is PennyLedger?

PennyLedger is an accounting system for clubs and societies, small business, and home use. As far as possible, the accounting entries are based on source documents. The source documents can be PDF's, emails or scanned paper documents.

PennyLedger is a client-server application. The server is a simple web server (written in Java and related technologies). The client can be any modern web browser.

PennyLedger is still under development. It has not been finished yet, but work is proceeding at pace.

In more detail

PennyLedger tries to do as much as possible to capture accounting information without configuration. The steps are:

  1. Source documents are imported. Source documents can be PDF's, scanned images or text files. Other document types could also be supported. The "import" can be done by uploading a file using a web client, or dropping the document into a "hot" folder. PennyLedger watches "hot" folders and imports any documents it finds there. PennyLedger can also watch email folders and import attached documents from there.

  2. All the text is extracted from the from the PDF's and OCR scanned images. If a PDF contains a background image, or consists entirely of scanned images, these are also OCR'd and the text extracted. The Tesseract OCR engine is used by default, but other OCR engines can be substituted.

  3. Machine learning techniques are used to classify the imported document. It uses the Weka Bayesian Network classifier to determine what type of document is being imported (invoice, statement, payment advice, etc), and what business entity produced the document (in Australia: Telstra, Optus, Commonwealth Bank, etc).

    This works well, even when there OCR is not entirely accurate. When the user reviews the imported document, if the classification is wrong, they have the opportunity to correct it. The corrected classification is then fed back into the Weka classifier so that when another document is received it is more likely to be correctly classified.

  4. Machine learning techniques are also used to identify significant fields of the imported document. For example, the significant fields of an invoice are: invoice date, amount, reference number and a short description of what the invoice is for.

    The Hidden Markov Model (HMM) is used to identify these fields. If the identification of the fields is wrong, the user can correct it. The corrected identification is fed back into the HMM so that when another dividend statement is looked at, it is more likely that the fields will be correctly identified.

    This step works well for documents where the text is cleanly read. That is, for PDFs or when the OCR of scanned documents is accurate. It is possible to enter the information required, but the philosophy of this project is to minimise such data entry.

  5. The significant fields are fed into a simple accounting system. A document generates one or more accounting transactions, which flows onto different accounting sub-systems, such as debtors and creditors ledgers, and the final general ledger.

    Across the world there are different rules for accounting systems. For example, value added tax (VAT) in the UK is called a goods and service tax (GST) in Australia and New Zealand, and the rules are different in all countries. This diversity is handled by using OSGi and having different code bundles, as necessary, to accommodate different accounting rules. A single set of code bundles can be used when all the source documents originate from a single country, or they can be combined to support multi-country accounting. While Java and OSGi supports internationalisation, PennyLedger have been developed using examples from UK, Australia and New Zealand.

Author

PennyLedger is being developed by Kevin Holloway. Kevin has 30+ years in IT, working as a Solutions Architect and Senior Developer. His most recent work has been on customer facing, web applications, using the full suite of Java technologies. He has a BA (Hons), a post graduate Diploma of Computer Science, and various industry qualifications.

Kevin now lives in his home town of Adelaide, Australia. Before that he was working in NZ and England.