Skip to contents

A dataset containing juxtaposing pairs of English words for 30 semantic relations. These anchors are used with the get_anchors() function, which can then be used with the get_direction() function. These have been collected from previously published articles and should be used as a starting point for defining a given relation in a word embedding model.

Format

A data frame with 303 rows and 4 variables.

Source

Curated from previously published semantic direction anchor sets. See: Kozlowski et al. (2019) "The Geometry of Culture"; Garg et al. (2018) "Word Embeddings Quantify 100 Years of Gender and Ethnic Stereotypes"; Bolukbasi et al. (2016) "Man is to Computer Programmer as Woman is to Homemaker".

Variables

Variables:

  • pole1. words to be added (or the positive direction)

  • pole2. words to be subtracted (or the negative direction)

  • relation. the relation to be extracted, 30 relations available

  • domain. 6 broader categories within which each relation falls