Document classification, data extraction and everything

No Comments

Over time, a lot of posts about document classification and data extraction, using Kofax, among other products, have been published in the codecentric blog. This blog post will put these posts into context and point out the changes with regard to older posts.

The codecentric divison ‘Digital Integration’ uses the products Kofax Capture / Kofax Transformation Modules and Kofax Total Agility, among others, in customer projects. Therefore, a large part of the posts refer to these products.

The listed posts were published independently over the last years. I have grouped the posts into different areas to get some clearity:

  • Best practices
  • Experience reports
  • Tips and tricks
  • Latest trends
  • The basis of everything
  • Best practices

Best practices in document classification and data extraction

Regardless of specific projects, these posts are about best practices in customer projects that use document classification and data extraction.

The basis of mailroom automation is the classification of incomig documents into different document types, as the extraction of data will be different within these types. The following article explains the classification tools available in Kofax Transformation Modules.

Document classification with Kofax Transformation Modules (KTM)

Artificial intelligence, neural networks and machine learning play a role in this environment. Kofax Transformation Modules offers AI mechanisms for years, providing new tools with every release:

Kofax Transformation Modules (KTM), AI and Machine Learning

Experience reports

The following blog posts are based on customer projects. Topics range from details of the processing of SEPA-mandates to document process auomation in insurance companies.

At one of our customers, incoming SEPA mandates are processed automatically or manually, depending on whether handwritten notes occur in a certain area of the SEPA form. This post explains how this can be done with KTM tools:

Kofax Transformation Modules: SEPA Mandates and handwritten additional information – or: who scribbled on my form?

Document process automation is discussed at the beginning of every project. At project start there often are different understandings about this topic and project members have to develop a common understanding about it before a project starts. These different views are presented in the following blog post:

KTM and insurance companies: Document Process Automation

The aim of a recognition process is preferably the automatic processing of incoming documents. Contract terminations offer potential for automation, as the termination date is often mentionend in the document. Practical problems may occur, but these can be solved with KTM tools:

Automatic termination of an insurance contract – how Kofax KTM may help

Tips and tricks

“Small” problems will occur in every customer project that cannot be solved with standard tools. This requires some creativity to find a solution without using external tools. Here are some tips and tricks which arose from our projects:

Scan and recognition products often try to rotate captured documents in a “correct” way, so people may read it without rotating the page manually. Sometimes this automation fails, especially with telefaxes, which may contain lines that are printed 90° or 180° rotated to the main text. The following article explains how to rotate this problem documents ‘correct’ automatically:

Orientation problems with document processing (Kofax Transformation Modules)

KTM offers so-called ‘dictionaries’. For example, you may use regular expressions for extracting a date from a document which may appear in different formats: 01.09.2019, 01. September 2019, etc. A dictionary (a plain text file) can contain the names of the months and their abbreviations. This dictionary can be referenced in the regular expression. This saves a lot of typing efforts when defining the regular expressions, and on the other hand, you may change the dictionary without modifying your KTM project. All this is KTM standard functionality. But sometimes you would like to search for stuff in the dictionary by script. This can be done this way:

Kofax Transformation Modules (KTM) – Dictionaries: Search by script

The next piece of advice is not necessary any more by now and may only be used for KTM versions 5 or lower. Machine-written data can easily be extracted by freeform recognition. This wasn’t possible with handwritten data, as full page OCR engines were optimized to machine-written characters. The following post describes how to recognize hand-written data with freeform recognition tools. Kofax KTM 5.5 and higher offers a new full text OCR engine which extracts machine- and handwritten data on a page.

Kofax Transformation Modules (KTM): ‘free-form recognition’ for handwritten numbers

The all-purpose tool for data extraction with KTM is the so-called format locators. The following two blog posts are an introduction to how to use these freeform recognition tools:

Kofax Transformation Modules – format locators and dynamic regular expressions

Kofax Transformation Modules – format locators and dynamic regular expressions – Part 2

Latest trends

For years, Kofax Capture and Kofax Transformation Modules have been the basis of many capture projects and Kofax is market leader in this area. To be prepared for advanced requirements, Kofax offers a product called Kofax Total Agility (KTA). In simple terms: KTA contains the products Kofax Capture, Kofax Total Agility and Kofax Import Connector embedded in a flexible workflow engine. Daniel Brodka explains the extensive capabilities of KTA in this post:

Introduction of and first steps in Kofax Total Agility

A growing part of our business is the area of Robot Process Automation (RPA). Kofax provides the product Kapow as platform to process data from structured or unstructured databases, files, email systems, websites, portals and even legacy mainframe systems or terminal emulations. Kapow fits perfectly into the other existing Kofax products. Kofax Kapow has changed its name recently and is now named Kofax RPA. Stefan Blank has summarized the capabilities of Kofax RPA/Kapow by building an example robot:

Robotic Process Automation with Kofax Kapow™

The basis of everything

The successful capturing platform offered by Kofax is Kofax Capture. With Kofax Capture you may get nice solutions without even using KTM. How to do this and how to create your own extensions to the platform is shown by Stefan Blank in this post about extended customizing of Kofax Capture:

Kofax Capture – Customisation beyond standard features

Stefan Blank wrote another blog post about a project-specific extension to Kofax Capture. This post is about adjustments of the scanning module to meet specific project requirements:

Kofax Capture Advanced Scan Api: A first approach

Kofax Capture includes a module to validate data that has been recognized or enter further data manually. This module is called ‘Validaton’. Within ‘Validation’, there is a scripting language which can be used to customize the behavior of the module to project requirements. For years, this language was ‘SBL’-Basic which is compatible to the ‘old’ Visual Basic of the 90s. But for some years it has also been possible to use .NET (VB, C#) as development environment. The next post explains what you need to consider when switching from SBL to .NET:

Kofax Capture Validation Scripting – from SBL to VB.NET for Dummies

Barcodes are a popular mechanism for document separation. The barcodes may be put as labels on the first document page or they may be inserted on a separate page before the first page of a document. The document separation works well in general. But sometimes external barcodes will generate wrong splitting of the documents, as they are recognized as separation barcodes. Now the document structure is destroyed. But there is a remedy for this:

Kofax Capture – Document Separation and Barcodes

Hopefully, this summary and sorting of the miscellaneous blog posts has made the topic of data capturing, data classification and data extraction more clear and transparent. For any questions and remarks please use the commet section below. We appreciate your feedback!

Tags

Comment

Your email address will not be published. Required fields are marked *