Code for the paper accepted at the 10th Workshop on Slavic Natural Language Processing 2025 (SlavicNLP 2025).

We conducted an analysis of semantic change in Croatian language over 25 years using word embeddings trained on 9.5 million news articles.

We investigate how word meanings evolve by training skip-gram embeddings on Croatian news articles from five-year periods (2000-2024). Our analysis captures linguistic shifts related to major events like COVID-19, EU accession, and technological changes. We also find evidence that embeddings from post-2020 encode increased positivity in sentiment analysis tasks.

We release the trained embeddings from five-year periods in this repository and also model trained on whole 25-year periods.

To obtain the embeddings:

1. Navigate to the 5ysplits_models folder
2. Run the data retrieval script:

   ./get_data.sh

This will download and set up all the trained embedding models.