A System for Affective Rating of Texts

In pursuit of automated text understanding, two broad types of approach can be distinguished: analytic methods (e.g. named entity extraction) that provide specific items of information, and synthetic methods (e.g. topic identification) that provide a global characterization. Recent interest in identifying overall affect or sentiment in text falls into the second category. Judging from the limited results reported so far, it appears to be a more challenging problem than topic identification. This is presumably because topic, to first approximation, can reasonably be represented by the straightforward accumulation of word content, whereas tone or affect -like meaning itself -depends on relationships of words with each other and with referents external to the text.