A text guided multimodal scale path fusion network for multimodal sentiment analysis
Siyuan Liu, Hongkun Zhao, Yang Chen, Fanmin Kong, Kang Li · 2025
Read A text guided multimodal scale path fusion network for multimodal sentiment analysis by Siyuan Liu, Hongkun Zhao, Yang Chen, Fanmin Kong, Kang Li (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/pmc-12827343.txt.
Description
Siyuan Liu, Hongkun Zhao, Yang Chen, Fanmin Kong, and Kang Li build a network for reading emotion across text, audio, and video in a paper from 2025. Older methods use fixed windows, which miss feelings that play out at different scales. Signals from voice and face can also clash with the words. Their network lets text guide the other channels. A transformer block scans features at several patch sizes and routes them adaptively.
A second module uses text-based memory to filter out conflicting audio and video signals. A channel fusion step then merges the streams for a final language model. On two standard data sets, the system beat most prior methods. The authors highlight its handling of mixed signals.
Catalog record
| CREATOR | Siyuan Liu, Hongkun Zhao, Yang Chen, Fanmin Kong, Kang Li |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 112 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |