In this study, we utilize a kernel theory of knowledge adoption model and propose a novel text analytic framework to classify the usefulness of solutions in online knowledge communities. The study combines multiple disciplines (behavioral, empirical, design science, and technical) to tackle an important and relevant business problem: how to effectively manage an online knowledge repository and identify useful solutions. Our framework can be implemented in online knowledge communities to improve users’ experience of searching for useful knowledge. The proposed framework has the potential to guide the development of customer-facing chatbots, which understand human-language questions and return helpful answers immediately. Online communities and social collaborative platforms have become an increasingly popular avenue for knowledge sharing and exchange. In these communities, users often engage in informal conversations responding to questions and answers, and over time, they produce a huge amount of highly unstructured and implicit knowledge. How to effectively manage the knowledge repository and identify useful solutions thus becomes a major challenge. In this study, we propose a novel text analytic framework to extract important features from online forums and apply them to classify the usefulness of a solution. Guided by the design science research paradigm, we utilize a kernel theory of the knowledge adoption model, which captures a rich set of argument quality and source credibility features as the predictors of information usefulness. We test our framework on two large-scale knowledge communities: the Apple Support Community and Oracle Community. Our extensive analysis and performance evaluation illustrate that the proposed framework is both effective and efficient in predicting the usefulness of solutions embedded in the knowledge repository. We highlight the theoretical implications of the study as well as the practical applications of the framework to other domains.