Fig. 5

Correspondence between topics extracted from unstructured written reports and the choice of predefined tags. A. Word clouds of the first five principal components for the cultivars Super Lemon Haze and Blueberry, indicating the most representative topics within the associated reports. B. Radar plots showing the mean semantic similarity between the words in each topic and the most frequently chosen effect tags for both cultivars. It can be seen that the semantic similarity is the highest for the most frequently used tags, and the lowest for a set of randomly chosen words unrelated to the effects of cannabis