Indico Data Solutions Inc., a provider of artificial intelligence-powered tools for process automation, today announced a major upgrade to its Unstructured Data Platform.
Indico has created a platform that’s used primarily by customers in the insurance, financial services and healthcare industries to automate the intake and analysis of documents and image-based data. The platform ingests PDF files, Word documents and various other unstructured data and processes them using AI tools such as optical character recognition.
That unstructured data can then be chained together into pipelines for data classification, extraction and comparison. It helps companies with tasks such as contract audits, customer onboarding, financial document analysis, mortgage processing and insurance claims.
The platform also makes it possible for developers to generate insights by overlaying additional machine learning models around sentiment analysis and keyword detection. It includes tools for developers to test and fine-tune those models.
In addition, the platform also uses a technique that’s known as transfer learning, where a model that’s tailored for one task can be used for another related task. Indico says it offers more than a dozen custom, out-of-the-box models that have been trained on datasets of more than 500 million documents. Customers can use these to analyze industry-specific data with just 200 training examples.
The latest version of Indico’s Unstructured Data Platform, Indico 5, has been designed to streamline what the company says are some of the toughest unstructured data automation problems of all: things such as document unbundling of PDFs and automating the human training corrections of AI models.
The headline feature in today’s release is the new automatic document unbundling capability, which allows AI to be trained to spit out documents from the most complex PDFs. Indico said this will be especially beneficial for processing mortgage applications and other financial paperwork that involves bundles of documents.
A second new feature, called Linked Labels, helps to eliminate the post-processing work required to reassemble relationships from extracted data, which it does by automatically capturing the relationships between document elements. That will eliminate the need for human workers to manually create labels for this type of data, Indico said.
Another update enables what Indico calls “staggered loop training,” helping to accelerate continuous improvement. So when humans correct data in the review phase, that data will arrive in the workflow already labeled, so the next version of the model reacts more easily and rapidly with fewer overheads.
Finally, there’s an enhanced optical character recognition model that can now recognize handwriting in documents as well as text. It’s a capability that should prove especially useful in healthcare, where millions of doctors still create handwritten notes.
Indico Data Chief Executive Tom Wilde said the latest version of the company’s platform is a major advance that will help make unstructured data more useful for thousands of business users.
“The real promise of automation in Indico 5 is using AI to augment human expertise, not replace it,” Wilde said. “The rapid evolution of workforce environments, where remote and hybrid working have shifted employee experience expectations, is also forcing businesses to rethink investments that improve accessibility and use of enterprise data, to increase productivity.”