February 12, 2019


The IRS expands its technological capabilities to help enforce tax law across 250 million Americans.

Since budget cuts in 2010, the IRS has been under significant pressure to be smarter in cases they choose to examine. According to Vidya Kauri of Law360, research and investigative tasks are now aided by machine learning technology. The IRS is deploying additional algorithms and artificial intelligence to mine data from tax returns, bank reports, property records and even social media posts to identify patterns in graphs where noncompliance might be present.

Additionally, in this high-performance computing environment, algorithms are used to fill in gaps in data or individual values on tax returns based on a pattern of surrounding data. The machine can also be working in the background looking for records that might be of interest while an agent looks at a case, and the agent can enhance the machine’s learning by letting it know if the records dug up are truly relevant or not.

The IRS is also using natural language processing technology that can enable a computer to read and to translate filings in foreign languages. Natural language processing and text analytics are also being used in appeal cases to try to understand why a particular case was lost or won, and then to predict the likelihood of winning, and on what founds, in other similar cases.

While the IRS seeks to maximize this technology to identify new ways to challenge tax returns, it is also helpful to millions of American who have been victims of tax account fraud. For example, a fraud technical advisor was able to identify 296 suspected identity theft returns claiming $1.3 million using such technology according to the IRS’ chief analytics officer, Benjamin Herndon.

Identifying anomalies and patterns as tax returns come in, enabling the IRS to run a number of iterations of a model or an algorithm, and applying natural language processing can now be done at a speed that was, before the last five to ten years, impossible according to Benjamin Herndon, IRS’ chief analytics officer.