Uncovering deep user context from blogs

Abstract

People’s utterances are fundamentally different to other documents because they are more immediate and less thought through. While this makes them more natural – noisy and unstructured – it provides an unrivalled opportunity to see “inside ” the author, to collect some context. The data requires analysis methods that have a relationship to human information processing: socio-cognitively motivated semantic systems. Using HAL, a method validated by cognitive science, the text from a large number of blog entries was analysed to uncover changes in entries author’s sense-of-self. Sense-of-self was measured through geometric projection of author’s first-person usage onto key indicators of kin and negative emotion words. An example of non-clinical qualitative evaluation affirmed the utility and promise of the technique: that deep personal context can be uncovered from utterances through the appropriate analysis and inference.