Research map / Architectures and efficiency / CNN design
ACT: an Attentive Convolutional Transformer for Efficient Text Classification
Pengfei Li, Peixiang Zhong, Kezhi Mao, Dongzhe Wang, Xuefeng Yang, Yunfeng Liu, Jianxiong Yin, Simon See
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use attentive convolution instead of conventional CNN feature maps (ablation) | ACT · Yelp F. development | accuracy (%): 67.1 -> 68.3 | Improved |
| Use attentive convolution instead of conventional CNN feature maps (ablation) | ACT · TACRED development | micro-averaged F1: 66.5 -> 67.8 | Improved |
| Add the multi-head structure (ablation) | ACT · Yelp F. development | accuracy (%): 67.0 -> 68.3 | Improved |
| Add the multi-head structure (ablation) | ACT · TACRED development | micro-averaged F1: 65.9 -> 67.8 | Improved |
| Include global representation in global attention (ablation) | ACT · Yelp F. development | accuracy (%): 67.6 -> 68.3 | Improved |
| Include global representation in global attention (ablation) | ACT · TACRED development | micro-averaged F1: 67.1 -> 67.8 | Improved |
| Include position embeddings in global attention (ablation) | ACT · Yelp F. development | accuracy (%): 67.4 -> 68.3 | Improved |
| Include position embeddings in global attention (ablation) | ACT · TACRED development | micro-averaged F1: 63.5 -> 67.8 | Improved |
| Use ACT instead of Transformer | text classification models · Yelp F. test | accuracy (%): 65.34* -> 68.16 | Improved |
| Use ACT instead of Transformer | text classification models · AGNews test | accuracy (%): 93.89* -> 94.25 | Improved |
| Use ACT instead of Transformer | text classification models · DBPedia test | accuracy (%): 98.98* -> 99.19 | Improved |
| Use ACT instead of Transformer | text classification models · Yelp P. test | accuracy (%): 96.13* -> 97.41 | Improved |
Closest papers by meaning
- Active Discriminative Text Representation Learning (AAAI 2017 · similarity 0.85)
- Graph Convolutional Networks for Text Classification (AAAI 2019 · similarity 0.85)
- An End-to-End TextSpotter With Explicit Alignment and Attention (CVPR 2018 · similarity 0.85)
- TextAdaIN: Paying Attention to Shortcut Learning in Text Recognizers (ECCV 2022 · similarity 0.84)
- On the Power of Convolution-Augmented Transformer (AAAI 2025 · similarity 0.84)
- Transformers Get Stable: An End-to-End Signal Propagation Theory for Language Models (ICML 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
