Various types of structured data collected by learning management systems such as Moodle have been used to improve student learning outcomes. Learning analytics refers to an assortment of data analysis methods used for this task. These methods typically do not consider unstructured data such as blogs, discussions, e-mail, and course messages. However, unstructured data analysis may reveal more effective indicators for improving learning outcomes. In addition, computing technology-enabled personalized learning also has great potential to enhance learning outcomes. This article describes an architecture for cognitive analytics and describes how it can be used to improve student learning outcomes through personalized learning. The author's approach is technology-enabled and data-driven. It draws upon the advances in cognitive computing for analyzing unstructured data to gain insights into student learning at the individual level. It employs automated question generation and assessment, and provides contextualized feedback. The DITA standard is used for authoring teaching, learning, and assessment materials. The approach integrates with and complements learning analytics.