Efficient Topic Extraction via Graph-Based Labeling: A Lightweight Alternative to Deep Models
2025
Read Efficient Topic Extraction via Graph-Based Labeling: A Lightweight Alternative to Deep Models (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-2511-04248.txt.
Description
Salma Mekaoui is the lead author among seven authors proposing a graph-based method to label topics extracted from text, as a lightweight alternative to deep models. Topic modelling finds themes as word distributions, which are statistically useful but often hard for a person to interpret without a clear label.
Topic words are enriched with semantically related terms, and relationships among them are explored as a graph. Labels derived from that structure aim to capture each topic's meaning accurately. In comparisons on two datasets against several benchmarks, including a commercial chat model, the graph method achieved consistently better labelling. The pitch is practical: good labels without heavy computation as text collections keep growing.
Catalog record
| YEAR | 2025 |
|---|---|
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 34 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-04248