When you look at the history of document capture (now being called Intelligent Document Processing by some), the last few years have been a wild ride. With the advent of advanced supercomputing, machine learning and artificial intelligence have been claimed and touted by all, especially newcomers in the industry. Here is a quick history in the advent of these industry-changing technologies, how the automation tiers are laid out, and the rise of true content understanding.

Old School

Quite a bit of document capture technology has been around for decades and is still the foundation of many legacy capture vendors that have not advanced their platforms.

Manual Classification and Data Entry

Ah, the old standby. You would be amazed how many organizations still manually open documents, manually identify them and manually enter data. It is the lowest on the automation totem pole, but it works (albeit very slowly).

Barcoding

Next up the chain is the use of barcode technology on documents to automate inbound processing and handling. Encode the type of content and the data that lies within, and when read, data is mapped to fields and extracted automated. The arrival on the scene of 2D barcodes like QR Codes, Datamatrix and PDF417 changed the game and allowed huge data sets to be held in a thumbnail patch.

Zone OCR 

With structured, repetitive documents or forms that have the same data in the same location consistently, “zoning” or Zone Optical Character Recognition (OCR) allows for automated extraction of data based on location and the conversion of image to text.

Auto-Classification and Text Matching  

Combining OCR with advanced pattern matching allowed large rule sets to be created to search for a certain text string and combinations of anchor values and text to not only classify the document but also to look for key information, regardless of the page position. This also allowed for broad variation in document structure and layout.

The New Kids

As computing power and technology have accelerated, the move towards modern data science and models have spurred organizations in a race for “document intelligence.” Now, it’s much more than just documents that have the need for intelligence – it’s all types of content from images, documents, emails, PDFs and videos. 

Machine Learning 

Training a system to do your document tasks for you seems to be the ultimate. Whether it’s an administrator or a consultant loading sample images, or an end user telling the system where the necessary data lies, machine learning is the technology of the day. Building a model to process documents seems the norm, but many systems fall short and lack granularity of control and operate as a black box.  

Dimensional Deep Learning and NLP

Let’s face it, documents are tough. Think about how humans interpret documents: we use our experience, text-based cues, understanding of business language and interpretation of layout and many other dimensions to classify a document to understand what is of interest. There are very few companies just touching on this space, using a broad range of dimensional analysis. This plus natural language processing (NLP) is the “tech du jour” in the intelligent processing sprint to the finish line.

The Future and What’s Next? 

Where will the industry go next? What’s on the horizon? Is there a game changer waiting in the wings? I am predicting that by year’s end there will be some grand announcements with some groundbreaking technology that goes beyond those in the race today. Keep moving forward with your intelligent digital transformation efforts, but don’t stop there: stay tuned for the latest updates as technology advances.

If you don’t follow us on LinkedIn and Twitter, make sure you do so today to keep track of industry breaking news.