Exploring Improvement on User Experience by Exploiting Patterns inside Crowdsourced Time-Sync Comments

Time-Sync Comment (TSC) is a type of crowdsourced audience review embedded in online video websites, and it provides better real-time user interaction than traditional user comments. Considering strong temporal feature of TSC, traditional approaches on analyzing textual comments cannot be transfered directly to solving TSC-related research problems, and hence it gains increasing attention in the past few years. However, there are three major problems on existing TSC research. First, they did not show usefulness of TSC compared to traditional information in improving user experience. Second, experiments were conducted on inconsistent TSC datasets, so it is hard to reproduce their results; Third, performance of existing methods is not convincible because their results were manually evaluated by limited number of so-called experts in their experiments. This paper aims to explore the usefulness of TSC data in improving user experience for watching videos by exploiting a larger-scale TSC dataset with four-level structure and richer attributes. Meanwhile, a set of TSC-related research problems are well defined in the paper and are solved by adapted state-of-the-art methods and evaluated based on crowdsourced labels contained in the dataset.