Men and women in discourse: An experiment on gender identification in discourse

Treball Final de Grau en Estudis Anglesos. Codi: EA0938. Curs acadèmic 2014-2015Diverse studies on how men and women make use of discourse (Lakoff, 1975; Coates, 1993) have obtained certain features that appear to characterise male and female texts. Nevertheless, it seems difficult to find a system which allows an accurate prediction of this pattern of use. There has been some attempts to create systems with those characteristics by means of computer programs such as the Gender Guesser, which was based on Koppel et al. (2002) and Argamon et al. (2003). Despite this, the degree of success of these computer systems is yet to be established. An aim of this study is to check if the Gender Guesser is a better predictor than English Studies Degree students with an average level of English equivalent to C1, as well as some knowledge about gender and discourse obtained in the degree, making them comparable competitors of the Gender Guesser in order to measure the success of their predictions. A comparison of the guesses on the gender of the author between a computer system named Gender Guesser and students of the last year of English Studies Degree was carried out. In addition to this, qualitative information regarding their perceptions was acquired. Results indicate that the computer system, which bases its analysis on the repetition of features from previous studies such as personal pronouns and specifiers, fails to guess most of the texts. On the contrary, the students, who base their answers on different features, are right in most of their judgements, obtaining higher accuracy rates than the Gender Guesser. Therefore, the difference between male and female discourse seems to lay elsewhere more personal and intricate than in the repetition of certain features. Further study on these new features may provide a better method to approach the prediction of male and female discourse production