Automated Interviews` Sentiment and Topic Analysis using BERT and TOP2VEC

A machine learning pipeline was developed to analyze the emotional tone, sentiment, and topics present in post-experiment interview transcripts related to out-of-body experiences (OBEs), in collaboration with the LNCO Lab at EPFL. Pre-trained models (VADER, BERT, Top2Vec) were used to perform group and individual-level analyses, identifying emotional and thematic differences between control and intervention conditions. Emotional synchrony between interviewer and participant was uncovered, along with the influence of experimental setup and interviewer bias on discourse patterns.