Autonomous Self-Assessment of Autocorrections: Exploring Text Message Dialogues

Text input aids such as automatic correction systems play an increasingly important role in facilitating fast text entry and efficient communication between text message users. Although these tools are beneficial when they work correctly, they can cause significant communication problems when they fail. To improve its autocorrection performance, it is important for the system to have the capability to assess its own performance and learn from its mistakes. To address this, this paper presents a novel task of self-assessment of autocorrection performance based on interactions between text message users. As part of this investigation, we collected a dataset of autocorrection mistakes from true text message users and experimented with a rich set of features in our self-assessment task. Our experimental results indicate that there are salient cues from the text message discourse that allow systems to assess their own behaviors with high precision.