A shopper passes by bags of rice at a supermarket in Seoul on Oct. 3, 2022. (Yonhap)
SEOUL, Oct. 18 (Korea Bizwire) — Artificial intelligence (AI) technology that analyzes new articles related to inflation will be capable of predicting the rising trend of consumer prices in the near future, a central bank report said Monday.
The Bank of Korea used an AI language model to measure the “written tone” of news articles dealing with inflation to see if it could be used to predict changes in consumer prices.
The model collected 64.06 million sentences from 188 new articles that dealt with consumer prices in the last 20 years.
Then, a human researcher divided 5,000 sentences into three categories — increase, no change, and drop in prices — depending on the “written tone” of each sentence, based on which the AI was trained to calculate the tonal index of all articles on inflation.
The results showed that the high and low mark in the tonal index preceded the high and low mark of the inflation rate by one or two annual quarters.
The tonal index deals with news articles that use all kinds of keywords related to consumer prices, which allows for a more precise and accurate prediction of inflation than analyzing singular variables like the producer price index or the breakeven inflation rate.
Kevin Lee (kevinlee@koreabizwire.com)
