Measuring sexually explicit content

Darroch, Katherine and Weir, George (2014) Measuring sexually explicit content. In: Cyberforensics 2014 - International Conference on Cybercrime, Security & Digital Forensics, 2014-06-23 - 2014-06-24, University of Strathclyde.

[img]
Preview
PDF (DarrochWeir-CFC2014-measuring-sexually-explicit-content)
10_darroch_weir.pdf
Preprint
License: Creative Commons Attribution-NonCommercial 4.0 logo

Download (871kB)| Preview

    Abstract

    In this paper we describe an experiment to investigate methods of measuring sexually explicit content in text documents. As a starting point, sample data was collected from a variety of sources and manually sorted into three categories: (i) sexually explicit, (ii) non-sexually explicit, and (iii) content that contained sexually explicit terms but was not sexually explicit - e.g., information used for sex education. This selection of data was used as a training set in developing three software metrics of the type often used in content filtering. Thereafter, a test set of six files was used in the experiment. These test files were scored for sexually explicit content by participants in the study and by the three different metrics. The assigned scores were then compared to consider how far the metrics and the users agreed on their view of sexually explicit content. In addition to our contrast between software metrics and users, we also note interesting trends from the participant demographics.