Multi-dimensional text feature fusion-based BA-RILA for ancient Chinese poetry theme recognition
Xinlu Zhang, Yuemin Liu · 2026
Read Multi-dimensional text feature fusion-based BA-RILA for ancient Chinese poetry theme recognition by Xinlu Zhang, Yuemin Liu (2026) 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/pmc-13219516.txt.
Description
Xinlu Zhang and Yuemin Liu build a neural system recognizing themes of ancient Chinese poems in a 2026 paper. Digitizing classical culture needs machines to sort poems by theme, but ancient Chinese lacks corpora and language models adapt poorly to its allusive style. Their framework, BA-RILA, attacks it by listening to a poem on three channels.
The system extracts vectors from a BERT pre-trained on ancient Chinese, rhythm features for the metrical pattern in eleven dimensions, and seventy-five features for its symbols and scenes. Attention weighting blends the three streams before a two-layer bidirectional LSTM reads the fused sequence. A multi-head attention stage sharpens the telling elements, and a classifier assigns one of six themes. The method outperformed benchmark models and generalized across Tang and Song dynasty sets. The authors present feature fusion, rather than a language model, as the way forward for text tasks.
Catalog record
| CREATOR | Xinlu Zhang, Yuemin Liu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 57 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |