On cleaning and organizing context logs for mobile user profiling

Mining data generated by mobile devices has drawn significant research attention in recent years. In this study, we investigate using text viewed, clicked, or entered into a mobile device to discover user preferences. We refer such I/O text as context text data, which will be a rich data source for bringing new opportunities for future mobile applications, such as understanding mobile users' preference and intention. However, unstructured context text data from various apps is with various information types and with noises. Aiming at this issue, in this study, we propose to organize the raw logs into behavior units based on text content similarity to address the issue of processing noisy, unstructured context logs. Experiments with data collected from real users are conducted to evaluate the performance of the proposed framework and the experiment results demonstrate the effectiveness of our framework.