Figure 3
From: Human and computer estimations of Predictability of words in written language
LMM analysis of the Gaze duration. (A) Different models includes the cloze-Predictability of the word N and different combinations of the NLP algorithms: 4-gram+cache, LSA (with a context of nine words) and FastText (with a context of 50 words). See Supplementary Table S1 for estimates and confidence intervals. (B) A new LMM with the cloze-Predictability was applied to the Gaze Duration with the fixed effects of each given model (columns) removed. (C) The difference between the AIC for each model was compared with the AIC of the baseline model (M0). Blueish colours represent negative t-values and reddish colours represent positive t-values. The exact t-values are also included in each cell. (D) p-values from ANOVA comparison between each model with the baseline model (M0.N) and with the cloze-Pred model (M1.N).
