Automatic gender detection of dream reports: A promising approach.

Abstract

A computer program was developed in an attempt to differentiate the dreams of males from females. Hypothesized gender predictors were based on previous literature concerning both dream content and written language features. Dream reports from home-collected dream diaries of 100 male (144 dreams) and 100 female (144 dreams) adolescent Anglophones were matched for equal length. They were first scored with the Hall and Van de Castle (HVDC) scales and quantified using DreamSAT. Two male and two female undergraduate students were asked to read all dreams and predict the dreamer's gender. They averaged a pairwise percent correct gender prediction of 75.8% (κ=0.516), while the Automatic Analysis showed that the computer program's accuracy was 74.5% (κ=0.492), both of which were higher than chance of 50% (κ=0.00). The prediction levels were maintained when dreams containing obvious gender identifiers were eliminated and integration of HVDC scales did not improve prediction.

DOI: 10.1016/j.concog.2016.06.004

Cite this paper

@article{Wong2016AutomaticGD, title={Automatic gender detection of dream reports: A promising approach.}, author={Christina Wong and Reza Mossanen Amini and Joseph M DE Koninck}, journal={Consciousness and cognition}, year={2016}, volume={44}, pages={20-28} }