Students of Informatics and Econometrics, Mikołaj Tym and Jakub Żerebecki, participated in the international scientific conference “Using Data from the Web to Shape the Next Generation of Labour Market and Skills Analysis”, organized by Cedefop at the headquarters of the European Economic and Social Committee (EESC) in Brussels.
The conference took place on April 1st, gathering around 60 participants — experts in labour market analysis using web data, representatives from European institutions, research centers, and statistical offices. Our students presented the results of their research showcasing innovative applications of language models in processing and analyzing job advertisements.
During their first presentation titled “Online job ad deduplication using large language models“, the authors introduced methods for identifying duplicate job postings published on European job portals. Their method detects four types of duplicates, including cross-language duplicates, using AI-generated semantic text representation. This allows for the removal of redundant content and ensures higher-quality labour market analysis statistics.
Their second presentation, “Online job ad classification using a fine-tuned encoder-based language model“, focused on classifying job ads according to the ISCO standard using a fine-tuned AI model. Accurate classification also required the use of an additional model to filter and eliminate irrelevant information. This method enables precise analysis of demand for specific occupations in the European labour market.
These studies represent an innovative approach to using artificial intelligence for public statistics in labour market analysis at a European scale.
More details about the conference are available on the organizer’s website: cedefop.europa.eu