Automatic Speech Recognition of Ukrainian Spontaneous Trauma-Related Speech: Accuracy, Linguistic Variability, and Evaluation Metrics. MIDI2025. Supplementary materials. Dataset.

The Excel file contains a dataset of AI and human transcriptions of audio-recorded Ukrainian narratives produced by 110 Ukrainian war migrants in Warsaw (Poland) about approximately 28 photos depicting Russian aggression in Ukraine and neutral images from OASIS in experiments, for evaluating the effectiveness of AI transcription.

  The dataset was developed as part of the research project "War mental health crisis coping: evidence-based media psychoeducation for family and community health promotion during economic recovery." A long-term program to support Ukrainian research teams at the Polish Academy of Sciences was carried out in collaboration with the U.S. National Academy of Sciences, with the support of external partners.