Gender composition of the documents.

In this paper we analyse gender-based biases in the language within complex legal judgments. Our aims are: (i) to determine the extent to which purported biases discussed in the literature by feminist legal scholars are identifiable from the language of legal judgments themselves, and (ii) to uncover new forms of bias represented in the data that may promote further analysis and interpretation of the functioning of the legal system. We consider a large set of 2530 judgments in family law in Australia over a 20 year period, examining the way that male and female parties to a case are spoken to and about, by male and female judges, in relation to their capacity to provide care for children subject to the decision. Structural topic modelling is used to develop coherent topics for sentences that fall under the notion of “capacity”, which are further differentiated by the gender of both the target of the sentence and the gender of the judge. The analysis reveals significant gendered differences in the language used in these documents, determined by both the gender of the target and the gender of the judge.