Extended Data Table 2 Semantic domains and word labels

From: Semantic encoding during language comprehension at single-cell resolution

  1. Left, grouping of words into semantic domains based on their 300-dimensional embeddings was performed through a spherical clustering and silhouette criterion analysis (Fig. 1c,d). Here, the labels were selected based on the preponderance of words within each grouping and were ordered based on the proximity of each word to their respective cluster centroid. Words obtained from the linguistic materials given to the participants optimally grouped into nine putative semantic domains. Right, to further confirm the selectivity of neuronal response (Extended Data Fig. 2b), words that were not intuitively related in meaning prima facie were either moved or removed to produce the following refined semantic domains.
Back to article page