【阅读笔记】End-to-End Neural Ad-hoc Ranking with Kernel Pooling
Xiong, Chenyan, et al. “End-to-end neural ad-hoc ranking with kernel pooling.” Proceedings of the 40th International ACM SIGIR Conference on Research and Development in Information Retrieval. ACM, 2017.
https://github.com/AdeDZY/K-NRM
概述:构建文本与文本的相似矩阵,使用 RBF Kernel 进行 Kernel Pooling ,取 log 相加后接一个全链接进行分类二分类(相似与不相似)。模型简称 K-NRM(Kernel-based Neural Ranking Model )。
过程是比较简明的:
先把 term 映射为 Word Embedding:
t ⇒ v ⃗ t t\Rightarrow \vec{v}_t t⇒vt
再计算Translation Matrix:
M i j = c o s ( v ⃗ t i q , v ⃗ t j d ) M_{ij}=cos(\vec{v}_{t_i^q},\vec{v}_{t_j^d}) Mij=cos(vtiq,vtjd)
然后通过 RBF Kernel:
K k ( M i ) = ∑ j e x p ( − ( M i j − μ k ) 2 2 σ k 2 ) K_k(M_i)=\sum_jexp(-\frac{(M_{ij}-\mu_k)^2}{2\sigma_k^2}) Kk(Mi)=j∑exp(−2σ