Explainability and effectiveness are two key aspects for building recommender systems. Prior efforts mostly focus on incorporating side information to achieve better recommendation performance. However, these methods have some weaknesses: (1) prediction of neural network-based embedding methods are hard to explain and debug; (2) symbolic, graph-based approaches (e.g., meta path-based models) require manual efforts and domain knowledge to define patterns and rules, and ignore the item association types (e.g. substitutable and complementary). In this paper, we propose a novel joint learning framework to integrate induction of explainable rules from knowledge graph with construction of a rule-guided neural recommendation model. The framework encourages two modules to complement each other in generating effective and explainable recommendation: 1) inductive rules, mined from item-centric knowledge graphs, summarize common multi-hop relational patterns for inferring different item associations and provide human-readable explanation for model prediction; 2) recommendation module can be augmented by induced rules and thus have better generalization ability dealing with the cold-start issue. Extensive experiments1 show that our proposed method has achieved significant improvements in item recommendation over baselines on real-world datasets. Our model demonstrates robust performance over “noisy” item knowledge graphs, generated by linking item names to related entities.
Description
Jointly Learning Explainable Rules for Recommendation with Knowledge Graph
%0 Conference Paper
%1 Ma:2019:JLE:3308558.3313607
%A Ma, Weizhi
%A Zhang, Min
%A Cao, Yue
%A Jin, Woojeong
%A Wang, Chenyang
%A Liu, Yiqun
%A Ma, Shaoping
%A Ren, Xiang
%B The World Wide Web Conference
%C New York, NY, USA
%D 2019
%I ACM
%K explanation knowledge-graph recommender
%P 1210--1221
%R 10.1145/3308558.3313607
%T Jointly Learning Explainable Rules for Recommendation with Knowledge Graph
%U http://doi.acm.org/10.1145/3308558.3313607
%X Explainability and effectiveness are two key aspects for building recommender systems. Prior efforts mostly focus on incorporating side information to achieve better recommendation performance. However, these methods have some weaknesses: (1) prediction of neural network-based embedding methods are hard to explain and debug; (2) symbolic, graph-based approaches (e.g., meta path-based models) require manual efforts and domain knowledge to define patterns and rules, and ignore the item association types (e.g. substitutable and complementary). In this paper, we propose a novel joint learning framework to integrate induction of explainable rules from knowledge graph with construction of a rule-guided neural recommendation model. The framework encourages two modules to complement each other in generating effective and explainable recommendation: 1) inductive rules, mined from item-centric knowledge graphs, summarize common multi-hop relational patterns for inferring different item associations and provide human-readable explanation for model prediction; 2) recommendation module can be augmented by induced rules and thus have better generalization ability dealing with the cold-start issue. Extensive experiments1 show that our proposed method has achieved significant improvements in item recommendation over baselines on real-world datasets. Our model demonstrates robust performance over “noisy” item knowledge graphs, generated by linking item names to related entities.
%@ 978-1-4503-6674-8
@inproceedings{Ma:2019:JLE:3308558.3313607,
abstract = {Explainability and effectiveness are two key aspects for building recommender systems. Prior efforts mostly focus on incorporating side information to achieve better recommendation performance. However, these methods have some weaknesses: (1) prediction of neural network-based embedding methods are hard to explain and debug; (2) symbolic, graph-based approaches (e.g., meta path-based models) require manual efforts and domain knowledge to define patterns and rules, and ignore the item association types (e.g. substitutable and complementary). In this paper, we propose a novel joint learning framework to integrate induction of explainable rules from knowledge graph with construction of a rule-guided neural recommendation model. The framework encourages two modules to complement each other in generating effective and explainable recommendation: 1) inductive rules, mined from item-centric knowledge graphs, summarize common multi-hop relational patterns for inferring different item associations and provide human-readable explanation for model prediction; 2) recommendation module can be augmented by induced rules and thus have better generalization ability dealing with the cold-start issue. Extensive experiments1 show that our proposed method has achieved significant improvements in item recommendation over baselines on real-world datasets. Our model demonstrates robust performance over “noisy” item knowledge graphs, generated by linking item names to related entities.},
acmid = {3313607},
added-at = {2019-12-25T19:19:06.000+0100},
address = {New York, NY, USA},
author = {Ma, Weizhi and Zhang, Min and Cao, Yue and Jin, Woojeong and Wang, Chenyang and Liu, Yiqun and Ma, Shaoping and Ren, Xiang},
biburl = {https://www.bibsonomy.org/bibtex/29aef4b7144d0c7a207ea5f9a4d821d52/brusilovsky},
booktitle = {The World Wide Web Conference},
description = {Jointly Learning Explainable Rules for Recommendation with Knowledge Graph},
doi = {10.1145/3308558.3313607},
interhash = {8ccad91b4a2a5ecc818d3480b2b97bd5},
intrahash = {9aef4b7144d0c7a207ea5f9a4d821d52},
isbn = {978-1-4503-6674-8},
keywords = {explanation knowledge-graph recommender},
location = {San Francisco, CA, USA},
numpages = {12},
pages = {1210--1221},
publisher = {ACM},
series = {WWW '19},
timestamp = {2019-12-25T19:19:06.000+0100},
title = {Jointly Learning Explainable Rules for Recommendation with Knowledge Graph},
url = {http://doi.acm.org/10.1145/3308558.3313607},
year = 2019
}