In an online environment, a baseline attitude of an author of online content is determined. Based on the baseline attitude and a raw sentiment score for an instance of online content, an adjusted sentiment score for the online content instance is generated. A variance from the baseline attitude may be detected, based on the online content of the author. In response to the variance, a current mood of the author is determined and, using the current mood and the raw sentiment score, another adjusted sentiment score for the online content instance is generated. The baseline attitude of the author may be determined using one or more of an analysis of the online content instance, a demographic profile of the author, and a subject matter area of the online content instance. The detection of the variance from the baseline attitude may incorporate a frequency of instances of online content.