Floor plans are essential sources for morphological studies in architectural history, offering insight into design thinking while opening to quantitative analysis. Yet, key contextual dimensions still rely heavily on the researcher’s interpretation. This study proposes a hybrid method that integrates graph-based similarity measures with human judgment to evaluate typological similarities in floor plans. We demonstrate the approach using a dataset of Dutch cross buildings, a rarely examined building type that served as community care centres. Our findings indicate that computational algorithms can capture nuanced similarity that exceeds human perceptual precision, while also revealing outliers to serve particular historical interests. Additionally, a flow-based algorithm produced numerical results that align closely with the architect's typological interpretation based on circulation. Together, these results offer new ways to understand building types, suggesting that quantitative insights can substantiate historical arguments, prompt new research questions, and extend to other building categories shaped by distinct spatial orders and use patterns. ... Read more