文章目录
我的论文笔记一般讲的都是干货,把我认为最有价值的部分提炼出来,当然也希望能帮助到大家。
大家共勉
Fine-grained Interest Matching for Neural News Recommendation
ACL 2020
Main Idea
Instead of aggregating user’s all historical browsed news into a unified vector, we hierarchically construct multi-level representations for each news via stacked dilated convolutions
Model
Three major components:
-
news representation module
-
cross interaction module
to exploit and aggregate matching information from
each pair of news at each level of granularity -
prediction module
Framework
news representation module
HDC(hierarchical dilated convolution) encoder to learn representations of news from multiple semantic views
空洞卷积
应用于CV中的语义分割,可以扩大视野同时保持分辨率
- input word embedding sequence
x i x_i xi is a d-dimensioned word emb
- use HDC
to capture multi-grained semantic features
- output
The output of each stacked layer l is preserved as feature maps of the news text at a specific level of granularity
Suppose there are L layers stacked, the multi-grained news representations can be defined as [d0, d1, . . . , dL].
Cross Interaction Module
Given representations of the k-th browsed news:
and the candidate news:
to build a segment-segment matching matrix for each granularity:
- fuses all interaction matrices into a 3D mathching image Q
n: total number of browsed news
each pixel Q k , i , j Q_{k,i,j} Qk,i,j is defined as:
note: each pixel is a concatenated vector with L + 1 channels
- 3D CNN
- output integrated matching vector s u , c s_{u,c} su,c
predictor
Dataset
MSN click logs
- F I M f i r s t FIM_{first} FIMfirst: use CNN instead of HDC
- F I N l a s t FIN_{last} FINlast: only use the last layer of HDC