This document contains the corpus-based predictions of gender expectations of more than 20.000 actions that are represented in ConceptNet. Currently, there are two predictions: Twitter-based (twitter.txt) and Web-based (ukwac.txt). For details consult the references at the end.

Both files have the same format: Each line consists of 4 tab-seperated fields, action, male count, female count, normalized score. Male and female counts correspond to the number of male and female utterances of the action either in a Twitter-based or Web-based corpus. The normalized score is a z-normalization based on a binomial assumption of male and female utterances with p equal to ratio of overall male utterances to the all utterances.

References: 

A. Herdağdelen, M. Baroni (2011). Stereotypical gender actions can be extracted from Web text. Journal of the American Society for Information Science and Technology, Volume 62, Issue 9, pages 1741–1749.

A. Herdağdelen (2010). Gender differences in tweets: A corpus-
based analysis of Twitter. Turunç Workshop on Complex Systems,
30 August-01 September 2010, Marmaris, Turkey.