Making Sense from Data: Moving from OCR to Smart Data Extraction | HCLTech

Making Sense from Data: Moving from OCR to Smart Data Extraction

May 29, 2020
Sunny Singh


Sunny Singh
Senior Manager, Intelligent Automation, DPO, HCLTech
May 29, 2020

The technology landscape of the world is now moving toward more capabilities. The rise in demand for more intelligent and auto-adaptive platforms has made technology leaders shift their focus to building platforms that transform the through intelligence. The technology now mimics human intelligence and partakes in business decision-making as well. Such advancements cut across every industry including data capture platforms and optical character recognition (OCR).

Traditional OCR– Optical Character Recognition

Traditionally, optical character recognition was limited to digitize scanned documents and images to extract entire text. As the technology became more mature, OCR technology providers wrote a piece of program or modules on certain templates of documents to extract relevant information. This worked well for the processes where the underlying document quality was good and the template was not changing or was with low variance. However, with time, the document types became more complex and adding variations that the standard OCR solution was not capable enough to manage. It encountered many challenges which include:

  • A high degree of manual effort was required to cater to the variances.
  • Even a slight variation in the format/structure of document made modules fail to extract information.
  • There was no solution to classify the documents based on certain rules.
  • It was practically impossible to write a program/code for over 20,000 document variations.
  • OCR technology failed to digitize low DPI documents, new fonts, character types, noisy documents, and handwritten data.
  • Extraction did not work on semi-structured and unstructured data.

Demand for a Smart Data Capture

All these shortcomings led to augment the capability of OCR to add a layer of intelligence which made OCR more powerful, adaptive, and cognitive.

The advent of technologies like , Deep / and has made it possible to create advance OCR that is capable of reading and interpreting the contextual meaning of the content of the document and extract meaningful / useful information in structured format. Not only this, OCR can now be programmed without coding any algorithm for any template.

What does Smart Data Extraction entail?

The Smart Data Capture is now powerful to demonstrate intelligence closer to human-level performance. It can now be trained using supervised learning just like we train a new agent joining the data entry process. This makes it easier for any user without any technical background to train the platform to extract relevant information from the documents.

Following are some of the benefits and features of using a smart data extraction platform:

  • The platform automatically detects image distortions, skewness, and rotation, among other factors, and applies relevant techniques to overcome these challenges.
  • ICR is trainable by the end user. The error in digitization can be fixed by taking feedback in real time from the end user automatically.
  • Data extraction for new fields can be trained on the fly automatically as the underlying algorithm will interpret the pattern of information for that field.
  • It improves efficiency and accuracy over time automatically.
  • Apart from text, image localization and extraction can also be achieved to extract image objects such as stamps, signatures, face captures, and others.
  • There is less dependency on the platform provider as the end user is empowered to train the solution.
  • The platform can be scaled according to multiple document types. Multiple categories of documents can now be processed through a single platform, with each type having a separate solution.
  • It has the ability to read documents with unstructured content.
  • Information in low-quality documents can be extracted by enhancing the image quality before extraction.
  • The platform makes it easy to track progress and estimate the effort saved. It can be integrated with data visualization tools to create meaningful dashboards to track efficiency over time.
  • Information with handwritten text can be extracted with ease.
  • The platform is scalable to multiple languages as it automatically detects language in the relevant document and applies the linguistic OCR it requires automatically.
  • The platform offers structured output in a format that is desired by the downstream application.
  • The platform is hardware agnostic. It works equally well with on-premises and .

As the technology is advancing, so is the ability to solve business challenges. This unleashes more creative and out-of-the-box solutions which can be scaled to deliver benefits that cuts across industries and horizontals. Intelligent data capture is a breakthrough in document processing and data entry processes that yields speed, efficiency, and accuracy to the customer in the way they manage their data.

The Future of Smart Data Capture

The Smart Data Capture is now moving from plain text extraction toward image object detection, classification, and interpretation of text and images to create summaries using natural language generation techniques.

The Smart Data capture is now moving from plain text extraction towards image object detection , classification and interpreting text and images to create summary using natural language generation techniques.

The platforms are now being trained on millions of domain-specific information and natural language generation techniques to be able to understand the context of the content along with nuisances that differentiate it from non-relevant text. This use of natural language generation will make for more powerful platforms that vertically integrate in business processes to aid in decision-making.

For example, such platforms can be used to read annual financial statements of a company to generate the summary of the company’s financial position and make recommendations for investors to invest accordingly. These platforms can also be used to read medical reports of a patient along with their clinical history to summarize the prognosis and make apt recommendations. They can also be used to perform fraudulent checks on documents to identify any money laundering indicators. The list is endless, and every industry will benefit multifold as the decision-making can then happen in a matter of minutes.

It is, therefore, very important for every organization to invest in these smart technologies as they are becoming powerful and their progress is also exponential. Most of these platforms are going to disrupt the industry as soon as they break through challenges and hurdles to match human-level intelligence.

To know more, about how we can help click here!

Get HCLTech Insights and Updates delivered to your inbox

Artificial Intelligence
Business Transformation
Machine learning
Share On