TOOLBOX

Songlines leverages large language models to analyze geo-text data from participatory or online discourses for urban design. It includes a pipeline constructing knowledge graphs that assign nonspatial information to a spatial context (i.e., a sentiment or preference to a specific location or feature), and a label set connecting this information to design-relevant contexts. It maps conflicting and mutualistic views on urban problems, experiences of spaces, behavior in spaces, expectations and opinions for urban development, and can be used in processing large amounts of discussions into input for planners, designers, or computational models.