Project description
Novel machine learning engine extracts and analyses documents like humans do
Typically, companies use only a fraction of their available data, as most of it is unstructured and hidden in documents such as emails and PDFs. The manual extraction of such data is repetitive, time-consuming, and error prone. Turicode has developed a machine learning engine that can read – and understand – any document like humans do. Once documents are transformed into machine-readable data, companies can easily search, analyse, and process previously hidden data to find new insights. The EU-funded MINT.extract project has set out to revolutionise document processing with a machine learning system that can be applied to any document type in every language. Turicode is advancing the research of data extraction, and commercially enabling large B2B companies to automate document management.
Objective
Around 80% of relevant business data is unstructured. To make valuable information from documents available for further analysis, lots of resources are invested in repetitive, time-consuming, error-prone and costly manual work. Efficient alternative solutions could reduce by 90% the time and costs employed in such tasks by any business. Digitization, i.e. transformation of human-readable documents into a digital form, is among the most common factors driving digitalization and a fundamental pre-requisite for automated text and data analytics. Digitization in EU could add €2.5trillions to GDP in 2025
MINT.extract is a disruptive information retrieval engine that delivers incredibly advanced document analysis capabilities, thanks to our innovative own-developed purpose-built document query language and AI based learning system. Using methods of artificial intelligence to transform unstructured documents into structured representations (database, XML…) and to read document elements (text, images, tables) as a human would do, our technology goes beyond current template-based solutions by automating many routine business processes and enables big data by integrating data from documents. We aim to create a generic learning system that can be applied to a diverse set of document types (e.g. insurance policies, purchase orders...) and delivers fully automated results in a quality that is superior to current manual data extraction.
With MINT.extract we will help businesses to transform their documents to value: making valuable information accessible for everyone. For our company, Turicode. We estimate that 5 years after Phase 2 completion, MINT.extract will bring us additional revenues of €18,7M (x54 revenues of 2018), allowing us to hire 50 new employees and generate €8,25M accumulated profit, reaching a ROI of 3,13.
Fields of science
Programme(s)
Funding Scheme
SME-1 - SME instrument phase 1Coordinator
8406 WINTERTHUR
Switzerland
The organization defined itself as SME (small and medium-sized enterprise) at the time the Grant Agreement was signed.