Stick-Breaking Embedded Topic Model with Continuous Optimal Transport for Document Streams
2025
Read Stick-Breaking Embedded Topic Model with Continuous Optimal Transport for Document Streams (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-18786.txt.
Description
Federica Granese, Serena Villata, and Charles Bouveyron build a topic model that can follow a stream of documents as it flows, rather than waiting for a fixed collection. News and social text arrive continuously and their themes shift, yet online topic models have been studied far less than their offline counterparts. Their model extends the Embedded Topic Model, which represents topics and words in a shared space, to process text batch by batch.
Two ideas carry the design. A truncated stick-breaking construction lets the model infer for itself how many topics are active at each moment. And a merging strategy based on optimal transport, adapted to high-dimensional topic space, combines the models from successive batches without losing older themes. Tests on simulated streams favored the new model over baselines, and an extensive real test used news articles about the Russian-Ukrainian war across 2022 and 2023.
Catalog record
| YEAR | 2025 |
|---|---|
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 67 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-18786