Learn about technology
Learn about Optical Character Recognition (OCR)
Role: Optical Character Recognition (OCR) is an engine.
OCR is the electronic conversion of images of typed, handwritten or printed text into machine-encoded text - whether from a scanned document, a photo of a document or from subtitle text superimposed on an image (for example, from a television broadcast).
OCR of input documents is a separate step of the development process - usually, this is the conversion from PDF into HTML/XML documents. The input quality of original documents influences a lot of the output of this process
WorkFusion works with ABBYY's OCR engine, which is integrated into separate business processes and proposes an OOTB (out-of-the-box) solution.
Learn about Tagging over Document (ToD)
Role: Tagging over Document is an interface.
In the 10.x version of Intelligent Automation Cloud, the proposed natural workflow introduces the Tagging over Document (ToD) labeling experience. A worker views the original document with text values layered underneath. The scan is rotated and scaled, and the image is optimized where needed for convenience. Word boundaries are highlighted and a single value can be selected with one click. Working with a document is as easy as in any text editor. To learn more, refer to TOD - Tagging Over Document.
Compare these cases
Let's compare two methods of tagging and decide which one is superior.
OCR without ToD
In version 9.x, workers tagged the plain text of OCR results. It might look nothing like the original documents because all information about sizes, fonts, positions, and layouts was lost. The workers sometimes couldn't even recognize the documents they used to work with. It took significant effort to precisely identify the required values. This led to user mistakes and increased labeling time.
This process uses standard OCR XML output to split the text into words, calculates absolute positioning coordinates for each word and generates HTML where the original page is used as background while the HTML words are put as an overlay on top of the image.
OCR with ToD
One of the main stages of an automation use case development is dataset tagging. Preparing and labeling a high-quality training set is time-consuming and costly. It equals a human worker teaching a bot to mimic human work - that is, to recognize values such as invoice number and date. Workers are accustomed to the real-world document structure and therefore can perform swiftly and effectively on what's familiar to them: hard and digital copies, and scans.
10.x version of Intelligent Automation Cloud restores the natural workflow by introducing the Tagging over Document (ToD) labeling experience. A worker views the original document with text values layered underneath. The scan is rotated and scaled, and the image is optimized where needed for convenience. Word boundaries are highlighted and a single value can be selected with one click. Working with a document is as easy as in any text editor.
See Tagging over Document in action on the GIF below.
A Tagging over Document workflow allows the following actions:
- Tag: select and label values, chunks of sentences, or entire table rows and columns
- Search: find other occurrences of a certain value across the document
- Evaluate: check the tagged values; click on the extracted value to jump to it even if the document has many pages
- An extensive assessment shows the average tagging time decreases by 52%
Tagging over Document makes the user experience intuitive. It doubles the speed of document tagging and cuts the time needed to teach a bot, thus streamlining automation use case delivery.
tip
Go to TOD - Tagging Over Document to learn about an out-of-the-box Tagging over Document business process, which allows users to label the original document so that they do not have to work with OCR results.
Meet AutoML
Role: AutoML is about adaptation.
Now, onto the component that makes our solution intelligent.
Why do we need AutoML? The problem is that OCR is an engine and ToD is a new advanced tagging interface. However, if the engine is not capable of concrete tasks and fails to recognize important documents, even a great interface can't help. To solve this problem, we have a patented AutoML technology that helps adapt the automation process to concrete business cases.
There are two particular features which complement each other:
- AutoML can review large volumes of data and discover specific trends and patterns that would not be apparent to humans. AutoML algorithms are good at handling data that is multi-dimensional and multi-variant, and can do this in dynamic or uncertain environments.
- Also, AutoML algorithms are capable of self-improvement. As the algorithms gain experience with data, their results keep improving in accuracy and efficiency. This leads to better decisions.
tip
To learn more about AutoML, refer to What can AutoML do?
Move further
Next, we'll study the tagging process itself with a step-by-step instruction.