Information from auditors, including alerts on the impact of the pandemic on business performance, is a basic element in analyzing the annual accounts that are being presented in 2020.
IBM Watson’s natural language processing capacity allows to automate interpretation of the auditors’ comments to improve the predictive capacity of Iberinform’s default risk ratings.
Iberinform has announced that it is using the natural language processing capabilities of IBM Watson to improve the predictive capacity of its default risk ratings. Iberinform daily recalculates the default risk of the 3.3 million active companies and freelancers operating in Spain by continuously capturing public information from hundreds of sources. A very valuable source, but difficult to treat, are the audit reports that collect the non-binding opinion of the external auditors on the annual accounts, which will be analyzed using the capabilities of artificial intelligence.
Each year, some 2.4 million pages of audit reports are published in the Commercial Registries. In some cases, the auditor notes that there are relevant deviations in the preparation of the financial statements. The analysis of your comments can contribute decisive elements when evaluating the commercial risk of a company beyond the image that reflects the numerical information of the balance sheet or the income statement. However, these are complex texts, which do not follow a standardized structure and are served in image format, which made any automation of the analysis extremely difficult. The only way to assess its importance was its manual review by specialized analysts. Until now.
The start-up of the project occurs at a key moment to facilitate decision-making by Iberinform clients, who have analytical tools and online reports to assess the risk of non-payment of an individual company or of its entire portfolio of clients in companies around the world. In the current environment of economic uncertainty, as the accounts that are drawn up include periods affected by the pandemic, the audit reports must assess whether the impact on the performance of the company is correctly collected.
Being a complex artificial intelligence project, the solution being used includes various technologies hosted on the IBM cloud and IBM services: from optical recognition of comments in image format, to IBM Watson Knowledge Studio, which extracts relevant concepts and data to correlate them, and Watson Machine Learning, which allows the deployment of advanced analytics models on the correlated data and returns the value of this predictive variable that Iberinform incorporates in its business scoring model. A fundamental aspect of the project has been the training of the system, where the analysis of thousands of audits of historical series of the last three years has been carried out to identify the relevant data that correlates with the deterioration of the payment capacity of said companies.
“We have implemented a business solution based on the latest cognitive and artificial intelligence technology, applying it to highly complex unstructured texts, to enrich the information we provide to our clients and improve our predictive models. It is one of the most cutting-edge projects in this field and has allowed us to strengthen our technological and analytical capacity to generate new valuable data for the market,” says Ignacio Jiménez, General Director of Iberinform.
“We are very proud to have been able to help Iberinform to implement a pioneering technological solution in its industry, which allows it to obtain a 
competitive advantage over its competition,” says Patricia Pisonero, Insurance Industry Partner at IBM Services.
Source: Iberinform Press Release
