Artificial intelligence (AI) and machine learning are included in almost every 2017 trends article, and rightfully so. Much like electricity in the first half of the 20th century and the internet in the past 20 years, AI and machine learning are poised to transform our lives in ways which were unimaginable only 10 years ago. In its list of the top 10 strategic technology trends for 2017, Gartner writes, “Artificial Intelligence and machine learning have reached a critical tipping point and will increasingly augment and extend virtually every technology enabled service, thing or application. Creating intelligent systems that learn, adapt, and potentially act autonomously rather than simply execute predefined instructions is primarily battleground for technology vendors through at least 2020.”
But what are AI and machine learning, and what impact will these technologies have on our industry? A2iA recently took time to talk with ECM Connection to answer these questions and more.
Artificial intelligence (AI) and machine learning are included in almost every 2017 trends article, and rightfully so. Much like electricity in the first half of the 20th century and the internet in the past 20 years, AI and machine learning are poised to transform our lives in ways which were unimaginable only 10 years ago. In its list of the top 10 strategic technology trends for 2017, Gartner writes, “Artificial Intelligence and machine learning have reached a critical tipping point and will increasingly augment and extend virtually every technology enabled service, thing or application. Creating intelligent systems that learn, adapt, and potentially act autonomously rather than simply execute predefined instructions is primarily battleground for technology vendors through at least 2020.”
But what are AI and machine learning, and what impact will these technologies have on our industry? A2iA recently took time to talk with ECM Connection to answer these questions and more.
Q: Are artificial intelligence and machine learning the same thing, or are the two different technologies working in conjunction with each other?
A2iA: Artificial intelligence (AI) is the process of solving tasks humans can complete easily but are harder for computers. Machine learning is considered a subset of AI that refers to the methodologies that enable software to learn specific tasks without being explicitly programmed. This includes neural networks and deep learning.
Q: How are these technologies applied to automate complex and handwritten workflows compared to ‘traditional’ document workflows?
A2iA: These technologies typically work hand in hand to simplify complex workflows. Traditionally, document workflows have difficulties automating documents containing complex and handwritten information. With deep roots in artificial intelligence, machine learning, and neural networks, A2iA’s in-house research and development team brings innovative software toolkits to market that are dedicated to image analysis and data extraction from everyday paper documents and checks.
Taking a step beyond traditional OCR (Optical Character Recognition) and ICR (Intelligent Character Recognition) technologies that only process printed and hand printed characters, the AI and machine learning components of A2iA’s proprietary IWR (Intelligent Word Recognition) technology allows workflows to automate access to freeform cursive handwriting by recognizing entire words and strings of words.
At the core of all of A2iA’s products, AI and machine learning fundamentals drive IWR processes to convert written information into digital data. Starting with a scanned image of a document, A2iA’s IWR technology preprocesses the image of a document. Entire pages and individual handwritten fields are de-slanted and de-skewed to remove the variability from the individual handwriting styles and noise on the page. Specific algorithms may be use to segment entire pages or paragraphs into lines. The fields or lines are then directly recognized by A2iA’s Neural Nets. Optionally, vocabularies or more complex Language Models may be applied in order to enhance the recognition accuracy.
Recently, A2iA released a new toolkit based on many years of R&D, taking new methodology and a new approach to AI and machine learning. This new approach involves a powerful engine that utilizes recurrent neural network (RNN) based technology used for both printed and cursive text recognition, allowing full page printed and cursive text to be transformed into searchable and editable formats — without the use of a dictionary. This new approach returns a full-text transcription with little to no customization and delivers a simple integration that allows organizations complete control over the recognition settings and results.
Easily integrated into existing workflows, A2iA’s core technology allows the unprecedented ability to locate, capture, and classify even the most complex, unstructured, and difficult-to-manage data from all types of document and convert it into readily accessible, searchable, and reportable digital information.
Q: What is workflow in the context of machine learning? Is it as simple as linking specific tasks in order to be automated?
A2iA: A2iA’s intelligent engines are designed to be applied to real world workflows and work with real world documents. Our technology makes these processes simpler, cheaper, and faster. Specific to a2ia DocumentReader, intelligent classification can be performed. During the engine setup, before real production begins, images from the workflow are used to train the engines based on samples. Similarly to how a human would learn based on repetition and training, the software does the same using its AI and machine learning algorithms.
Q: What are the benefits of automating workflows utilizing technology comprising AI and machine learning?
A2iA: Organizations automating their document workflow processes utilizing AI and machine learning based technology stand to gain more control and more efficient accessibility to data regardless of the project or document type. The customization and scalability of A2iA’s technology allows organizations the ability to access complex information that would otherwise be inaccessible by traditional recognition technology. To put this more simply, A2iA’s engines offer organizations a vital way to improve their current workflows as well as increase their capabilities and expand their offerings in order to grow their business.
Q: How can AI/machine learning technology be effectively leveraged on mobile devices to automate workflows?
A2iA: In an increasingly mobile world, workflows continue to evolve forcing organizations to find new ways to streamline workflow processes and to include mobile images into their environment. The faster data is accessible throughout the enterprise, the higher value it has and the more it contributes. Effectively leveraging AI and machine learning technology to drive data accessibility is instrumental to advancing automation processes.
Organizations are faced with driving easier and more user friendly experiences than their customers have grown accustomed to. Mobile images can add additional complexities to the workflow, as the image quality is not always as guaranteed as it is from a desktop scanner, and the orientation and edges of the document need to be detected, cropped, and corrected before the image can be processed.
Embedded directly into mobile apps or device operating systems, A2iA’s mobile SDKs are tuned to do exactly this, and bring the same artificial intelligence and machine learning principles trusted in desktop applications directly to the mobile device. These engines are geared to capture and extract machine printed, hand printed, or cursive text from checks, forms, ID cards, and full documents without having a server, data or Wi-Fi connection, including the capture, image cleanup, and text recognition. Its technology allows mobile users a quick and simple user experience with reduced and sometimes even eliminated manual keying or interaction. Organizations can move more quickly and more effectively use data collected and can include all types of data into the workflow, regardless of the method of capture.
About A2iA
The AI and machine learning based recognition toolkits that A2iA brings to the market are cultivated by more than 40 years of research and development by the late Professor Jean-Claude Simon, a pioneer in handwriting recognition. Professor Simon wanted to transform the results of his team’s research into real world business applications and founded A2iA to do just this. Today our award winning toolkits are used worldwide to automate complex document processing workflows. By extracting more information from any digitized document regardless of its type, A2iA’s software offers companies the ability to gain more access to the handwritten unstructured data previously out-of-reach, turning paper document into meaningful information that can be integrated into companies’ information systems. Because of the flexible and scalable nature of the toolkits, A2iA’s software can be integrated into BPO platforms to enhance their processes and speed access to their customers’ data, as well as at organizations who strive to automate their internal workflows and paper-based processes.