Crises of reputation as asynchronous online polylogues

Crises of reputation are rarely studied from the linguistic point of view, which results in several research gaps. Therefore, in this study three selected crisis cases have been ana-lyzed as asynchronous online polylogues-Internet debates performed by stakeholders in different sources from the moment of publication. In this approach corpus analy-sis has been combined with the theory of computer-mediated communication, speech acts and a semantic study of emotions. One of the findings was that certain common patterns of subtopics, emotion expressions and communicative actions are noticeable in the user-generated content across different crisis cases. Contrary to some NLP stud-ies, emotions expressed by stakeholders in a crisis situation refer not to the major topic, but to its subtopics, which impacts the construction of text-mining models. Unlike in pre-social media studies, sarcasm and not anger is the major emotion expressed textu-ally in reaction to a crisis of reputation.