Fig. 8: Large-scale analysis of conversational data replicates the temporal bias in conversational references.
From: Temporal asymmetries in inferring unobserved past and future events
We used natural language processing to automatically identify references to past or future events across a variety of datasets. The “type” label was applied following19. For each dataset, we calculated the ratio of references to past events and references to future events. We performed a large-scale analysis of the ratios with a meta-analysis-like approach using a random effects model. Data are presented as ratios ± 95% CI, with a summary estimate based on the random effects model. Dataset descriptions may be found in Supplementary Table S3.
