OCR stands for Optical Character Recognition and is a technology that can “read” and convert text from the image of a document into digital text. It is based on Artificial Intelligence (AI) and machine learning and used to index, search, edit or use the text in machine processes.
Over the course of the last few years, we have witnessed a sizeable increase in the platforms, tools and applications making use of these technologies. Statista expects global revenue from the AI market to increase from 482 billion dollars in 2018 to 3,061 billion dollars by 2024. The neural network software market is expected to expand to over 22,5 billion dollars in the same period, at an annual growth rate of over 30%.
This is because AI applications have gone beyond the IT concept and influenced operations in healthcare, manufacturing and the legal industry among many others. AI-related initiatives are popping up across the board, as evidenced by the cooperation between industry giants AWS, Facebook and Microsoft to build the Open Neural Network Exchange (ONNX).
The AI market is expected to grow by over 500% in the next five years.
As a key member of the European FinTech market, MobileXpense has distilled the best of this technology to make it available to our valued customers. In 2017, we joined the OCR revolution and have since been working tirelessly to improve our offering. We want our customers to save more, optimize their efficiency, and focus on what matters - all with the help of OCR.
Earlier iterations of OCR software involve rules and templates to “teach” the software to capture and recognize data by comparing it to a stored glyph, pixel-by-pixel. This “pattern matching” can be performed by neural networks. However, this type of software works best for typewritten text and offers little flexibility.
Modern OCR software breaks up the glyphs into “features” including lines, loops, intersections and directionality. These features are then compared to those of stored glyphs to identify the nearest match. More advanced software even performs a second pass or “adaptive recognition”. This uses the shapes recognized in the first pass to increase recognition of the remaining shapes.
Machine learning and neural networks further help increase accuracy.
Machine learning further helps increase accuracy. It is an application of artificial intelligence in which a model obtains data that is then fed back into the model, improving it over time. By providing a “dictionary” from which the words are most likely to come, the accuracy of OCR is greatly improved. Neural networks perform these tasks thanks to their pattern recognition capacities and their ability to learn and make decisions.
OCR has matured out of the realm of the experimental and found real-world applications it excels in. Thanks to the power of the cloud, companies are leveraging machine learning to come up with smart and scalable solutions to existing challenges.
In banking, OCR was initially used to archive client data and process cheques with high accuracy and a reduced turnaround. It has now grown beyond this by helping to reduce instances of fraud; identifying forgery and spotting problematic information in credit card, loan or other applications. According to a report by Celent, machine learning is used to analyze credit applications that later defaulted to identify risk-predicting factors. It is further used to compare transactions to known proper and fraudulent transactions to assess whether the transaction should be cleared or flagged.
The banking, legal and healthcare sectors are taking OCR out of the experimental world.
In the legal as in the healthcare sectors, OCR is part of the drive towards paperless institutions. Digitized documents are much easier to index, store, search through and create automated reports from. Statements, wills and affidavits but also medical data are now rapidly digitized thanks to OCR.
As a reference for Travel and Expense Management (TEM) services in the FinTech sector, MobileXpense extracts the best of these technologies for the benefit of its customers.
After first making OCR available to our customers in 2017, we have constantly sought out new, better providers and sealed partnerships to help us harvest even more of the power of OCR and AI. In 2021, we aim to revolutionize travel and expense management with a user-centric approach that will help them save more time and money.
Most people type at speeds of between 35 and 40 words per minute, whereas OCR data entry is almost instantaneous. With our SpendCatcher app (App Store, Google Play), users can simply snap a picture of their receipts and forget about them. Our OCR solution extracts the payment information and creates a draft expense line in just a few seconds, savings our users on average 50% of their time spent entering expenses.
MobileXpense users spend 50% less time on their expenses.
With accuracy rates consistently hovering above the 99% margin, this means our users are also spending up to 95% less time controlling expense data previously captured by OCR thanks to our automated control tool SpendController. In comparison, Bloomberg stated in 2015 that up to 27,5% of U.S. businesses reported manual data entry errors, costing them billions in tax penalties.
The above translates to a 65% saving on processing costs by replacing many man-hours of data entry and control with OCR. The increased insight we provide our customers into their spending and travel also allows them to save an average of 15% on their TEM budgets.
The uses for AI and machine learning range far and wide and will keep growing in the near future, as made clear by the increasing number of companies dedicating resources to AI.
Our SpendCatcher app is one of the best examples of harnessing these technologies to help simplify people’s lives, allowing them to focus on what truly matters to them. While we do not believe that machines will be replacing humans in the near future, we are intent on using their capabilities to improve our users’ daily lives.