Topic Tracking from Classification Perspective: New Chinese Dataset and Novel Temporal Correlation Enhanced Model

Jin Ma, Xinming Zhang, Yuxun Fang, Xinyu Zuo, Haijin Liang

Abstract: Topic tracking focuses on linking novel events to existing topics and is critical for applications. In this work, a deep learning model called TDE is proposed to solve the topic-tracking problem. And the first industrial-level topic tracking dataset QBETTD with fine-grain score labeling is released. The TDE model topical correlations among events from semantic and temporal perspectives based on machine reading comprehension (MRC) structure, which captures the deep semantic meaning and inter-relationship of the events and topics. The empirical results show that the proposed method is suitable for the topic-tracking task, and the model can track a growing topic.

External IDs:dblp:conf/nlpcc/MaZFZL23