Reproduce “FE-TCM: Filter-Enhanced Transformer Click Model for Web Search” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yingfei Wang, Jianping Liu, Jian Wang, Xiaofeng Wang, Meng Wang, Xintao Chu
FE-TCM presents a filter-enhanced Transformer click model designed to predict user interactions in web search by mitigating noise in behavior data through learnable filter layers. The model achieved superior click prediction accuracy on real-world datasets, outperforming both probabilistic graphical and neural network-based baselines.
Key insights
- Learnable filter layers, adapted from digital signal processing, effectively attenuate noise in user behavior data, significantly enhancing the robustness and accuracy of deep click models.
- The Transformer architecture provides a powerful mechanism for extracting intricate dependencies from diverse features in click sequences, further optimized by noise filtering.
- The `exp_mul` combination function, which supports the examination hypothesis with learnable parameters, yields the most effective integration of examination and attractiveness predictions for overall click probability.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.