Answers are grounded in this paper and cite their sources.
Discourse analysis study reveals power dynamics in Indonesian graduate supervision dialogues using a hybrid BERT model, highlighting pathways toward empathetic and culturally aware mentorship.
The linguistic organization of advice-giving and the enactment of power in research supervision are critical yet often overlooked dimensions that shape students’ learning agency and well-being. The present study integrates Conversation Analysis (CA) with text mining using Bidirectional Encoder Representations from Transformers (BERT) to model how power relations are linguistically realized in Indonesian supervisory discourse. A corpus of 1,200 annotated utterances was collected from supervision sessions across five Indonesian universities, encompassing public and private, Islamic and non-Islamic institutions. The hybrid CA and BERT classifier achieved a weighted F1-score of 0.91 and an area under the receiver operating characteristic curve of approximately 0.94, outperforming the baseline Convolutional Neural Network and Support Vector Machine models. These demonstrate that integrating context-sensitive qualitative analysis with computational classification improves the accurate identification of power relations in supervisory interactions. These computational findings extend qualitative theories of supervision by providing scalable, evidence-based measures of power and relational empathy. The integration of CA and BERT methodologically contributes to the development of an empathic advice-giving model for inclusive higher education, demonstrating how text mining-assisted discourse analysis can promote reflective and culturally aware supervisory practices.
No takes yet. Share an insight, caveat, or question.
Nimasari et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: