MIT Press

Skip to Main Content

Table 9

Neural (S_token and S_type) logography measures computed on the subsets of Wikipedia using transformer architecture V. Multihead attention strategies similar to Table 8.

Language	V_A			V_M
Language	$S_{tok}^{V_{A}}$	$S_{typ}^{V_{A}}$	Acc.	$S_{tok}^{V_{M}}$	$S_{typ}^{V_{M}}$
English	0.53	0.42	0.94	0.53	0.42
Finnish	0.46	0.40	0.96	0.54	0.50
Japanese	0.86	0.75	0.92	0.84	0.73
Korean (jamo)	0.36	0.31	0.97	0.34	0.30

Language	V_A			V_M
Language	$S_{tok}^{V_{A}}$	$S_{typ}^{V_{A}}$	Acc.	$S_{tok}^{V_{M}}$	$S_{typ}^{V_{M}}$
English	0.53	0.42	0.94	0.53	0.42
Finnish	0.46	0.40	0.96	0.54	0.50
Japanese	0.86	0.75	0.92	0.84	0.73
Korean (jamo)	0.36	0.31	0.97	0.34	0.30