Recent years have seen increasing interest in analyzing the millions of texts published every year in blogs and forums. One of the applications attracting most attention is the automated gathering opinions of people on different products. In this paper, we elaborate a methodology for opinion mining and subjectivity analysis, based on the notion of personal sense. Our hypothesis is that the personal sense of the words that the author uses in a text is reflected their appraisal of a product described in the text. In this preliminary work, we describe a new methodology applied to the polarity classification task, and present and evaluate our the preliminary results.