English verb regularization in books and tweets
Fig 2
Comparison of verb regularization for Ngrams and Twitter.
We calculate verb regularization fractions using the datasets in row (I) of Table 1. Verbs are centered at their regularization fraction in Ngrams (horizontal) and Twitter (vertical). Both axes are on a logit scale, which spreads out both extremes of the interval (0, 1). Verbs to the right of the vertical dashed line are regular in Ngrams; verbs above the horizontal dashed line are regular on Twitter. The diagonal dashed line separates verbs that are more regular on Twitter (those above and to the left of the line) from those that are more regular in Ngrams (those below and to the right of the line). For example, compared with ‘knew’, the word ‘knowed’ appears roughly 3 times in 1000 in Ngrams, and 2 times in 10,000 on Twitter, making ‘know’ irregular in both cases, but more than an order of magnitude more regular in Ngrams than on Twitter.