Merging Embedded Topics with Optimal Transport for Online Topic Modeling on Data Streams
This work addresses the need for efficient topic modeling in dynamic text streams like social media, though it is incremental as it builds on the Embedded Topic Model.
The paper tackles the problem of online topic modeling for rapidly growing social media data streams by introducing StreamETM, which merges models from consecutive batches using unbalanced optimal transport and includes change point detection, resulting in outperformance over competitors in numerical experiments on simulated and real-world data.
Topic modeling is a key component in unsupervised learning, employed to identify topics within a corpus of textual data. The rapid growth of social media generates an ever-growing volume of textual data daily, making online topic modeling methods essential for managing these data streams that continuously arrive over time. This paper introduces a novel approach to online topic modeling named StreamETM. This approach builds on the Embedded Topic Model (ETM) to handle data streams by merging models learned on consecutive partial document batches using unbalanced optimal transport. Additionally, an online change point detection algorithm is employed to identify shifts in topics over time, enabling the identification of significant changes in the dynamics of text streams. Numerical experiments on simulated and real-world data show StreamETM outperforming competitors. We provide the code publicly available at https://github.com/fgranese/StreamETM.