| Popis: |
Exploration of real-time summarization (RTS) methodologies and applications to esports events on Twitter. The goal of this study is to evaluate the effectiveness of real-time summarization techniques at esports event detection, highlight summarization, and timeline generation. A two-step system of event-prediction and summarization is proposed. First, using Twitter as the data source, events in an esports game are predicted through machine-learning and classification to determine event occurrences. Four major classification features and five standard classification models (Naive Bayes, Logistic Regression, Decision Trees, K-Nearest Neighbors, Support Vector Machines) are evaluated to identify an optimal event-detection model. Second, natural-language text processing functions such as term-frequency and TF.IDF are evaluated for effective event summarization and to confirm successful event-detection. The CART (classification and regression tree) model is selected as the most optimal model for event-detection, predicting in-game esports events with 70% accuracy. This study demonstrates the application of Twitter as a data source in detecting real-time esports events. |
