No open-access location is known for this record yet.
No open-access location is known for this record yet.
Monitoring and understanding human migration as triggered by a crisis is challenging. Combining spatial analysis with natural language processing when analyzing social media data helps to understand the mobility and the needs of migrants better. For this paper, we used geolocated Twitter data to analyze the mobility of and topics discussed by migrants of the Ukraine war in 2022. We removed bots, accounts showing implausible mobility, and automated text content from our dataset. Then, we applied a transformer-based multilingual topic modeling framework to identify the migrants’ discourses. We assessed the topics discussed by migrants before leaving Ukraine, after leaving Ukraine and after returning to Ukraine. Our results show that “Attack reports”, “politics”, “donations to Ukrainians”, “food export/production”, “humanitarian aid”, “nuclear threat”, “Ukrainian places”, “job search”, and “war journalism” were dominant topics before leaving from and after returning to Ukraine. “Food”, “social media”, “transport”, “art”, and “finance”, however, were important topics right after leaving the country. Overall, our results reveal plausible spatial patterns of migration, which are similar to those reported by official statistics (R2 = 0.89), showing the reliability of geotagged social media data to monitor human mobility. This information can complement official sources, adding first-hand information on the mobility and needs of migrants across space, time, topics, and languages. This is crucial to develop humanitarian response plans when time is of the essence.
Loading…
Stored only when both papers are in the corpus — 20 citations exist upstream.
Papers that cite the same references — the relational fallback while embeddings are still backfilling.
Imported from OpenAlex (CC0) — nothing here is generated.