Abstract: The landscape in which students, alumni, and potential applicants continually
negotiate the institutions’ brand perceptions is through unstructured digital discourse. Spanning
47,832 social media posts and comments collected from ten public universities ac ross October–
December 2024, this study also triangulates a previously validated survey-based structural model of
consumer-based brand equity (CBBE) in Azerbaijani higher education (Paper 1) with a large-scale
social listening and natural language processing (NLP) analysis. A custom five stage NLP pipeline—
including AzerBERT sentiment classification (macro-F1 =. 881), aspect-based sentiment analysis
(ABSA; mean κ =. 776), BERTopic/LDA topic modelling and social network analysis was established
to overcome Azerbaijani agglutinative morphology and widespread code-switching. In a prespecified triangulation protocol, all five structural paths were validated and directionally consistent.
Three findings are theoretically significant: scholarship−delay discourse (1 8.4% negative posts;
polarity −0.67) pinpoints the mechanism of brand credibility-trust hyperscale; urban-non-urbanclustering coefficient differentials (0.671 vs 0.412; d = 2.69) identify location -moderated equity
conversion; and platform-specific thematic divergence (χ² = 847.3, V =. 42) illustrates differential
brand-dimension activation across channels.
Keywords: social listening; natural language processing; brand equity; higher education;
AzerBERT; aspect-based sentiment analysis; BERTopic; social network analysis; Azerbaijan, mixedmethods triangulation.