Monday, July 18, 2005
Contextuality of Blogs, 3 - Individual Differences
In the last post I talked about how a measure of formality/contextuality could be used to tell the difference between different genres. It does this by looking at one aspect of the language they use, namely the parts-of-speech. Here I will talk about the work I've done on looking at individual differences: looking at gender and personality.
There has been a lot of work on gender and language. Perhaps the most well known application is the much linked gender genie, based on the work of Koppel and Argamon. I used the F-measure to see if there was any difference in formality between men and women. Not only was there author gender information for our blogs and emails but the BNC held information for some of its genres. You can see the resulting F-scores below.
In four of the five genres women score significantly lower than men. As we might expect, men take a more formal approach to writing, while women are far more contextual. The exception is academic writing. Here, the almost identical levels of formality suggest that when required, women can adopt a style as formal as that projected by men.
The next investigation concerns the personality of the bloggers in my study. As part of my data gathering, bloggers completed a personality questionnaire that gives scores on 5 factors: Neuroticism, Extraversion, Openness, Agreeableness and Conscientiousness. For an explanation of these traits scroll down this page.
To investigate how formality/contextuality differs by personality, we correlated each trait with the authors' f-scores. The previous work on implicitness in emails leads us to expect high neurotics and extraverts will prefer contextual language. The results of the correlations can be seen in the next table.
The negative score for Extraversion and Neuroticism is as we expected: higher scores on the personality scale, relate to low F-scores and contextual language. Low neurotics and introverts write in a more formal style, but only just. The correlations are actually very low, so the results are far from conclusive.
Conscientiousness has a very low score, meaning there is no correlation at all. Openness and Agreeableness however score much higher correlations, the latter being a statistically significant result. Openness has previously been considered the factor of intellect, and it was theorised that higher scores on the Openness trait would reflect higher levels of formality. Our result show some support for that theory.
More interesting is the novel Agreeableness result, for which we have our own theory. One aspect of Agreeableness is the cooperative and accommodating nature of high scoring individuals. This suggests to us that highly Agreeable bloggers are more aware that there may be little context shared between them and their readers. This results in a much less contextual (more formal) style of writing than low Agreeable bloggers.
This is by no means the end of my work, but it sums up everything that I'm taking to the conference. I've introduced you to the F-measure, shown you that it can show differences in genres, and it can highlight differences between individuals. In the future, I will return to formality/contextuality, as I have discovered a few more interesting things, but I'll leave it here for now.
If you would like more details on this work and the background to it, you can always check out the original paper. If there is anything else you would like to know, if you have any questions or want leave any feedback, then you can always leave a comment here or alternatively drop me a line.