BERLIN, April 16 (Reuters) - The use of artificial intelligence will increase accuracy when predicting the monetary policy moves of the European Central Bank, a study by the German Institute for Economic Research DIW Berlin said on Wednesday.

Researchers analysed the ECB's communications from January 2019 to March 2025 using a specially trained text analysis model based on AI.

The program takes each sentence of the ECB's statements individually and analyses whether it is a signal for a restrictive, expansionary or neutral monetary policy, DIW expert Kerstin Bernoth said.

In an extended forecasting model - using the text analysis and taking into account inflation, economic policy uncertainty and the previous interest rate course - the accuracy of forecasts for interest rate changes can be increased from around 70% to 80%, the study showed.

"Central banks use language as a monetary policy instrument," said Bernoth, author of the study. "The choice of words in speeches, press releases or interviews is never random, but carefully considered and allows conclusions to be drawn about the future direction of monetary policy."

For the upcoming ECB meeting on Thursday, the forecast model signals a high probability of a further interest rate cut, despite the recent more neutral tone.

Analysts polled by Reuters expect the ECB to cut its key interest rate from 2.5% to 2.25% as tariffs curb trade and uncertainty weighs on consumption and investment.

Reporting by Rene Wagner and Maria Martinez, Editing by Rachel More and Ed Osmond