Learning What You Need from What You Did: Product Taxonomy Expansion with User Behaviors Supervision

Abstract

Taxonomies have been widely used in various domains to underpin numerous applications. Specially, product taxonomies serve an essential role in the e-commerce domain for the recommendation, browsing, and query understanding. However, taxonomies need to constantly capture the newly emerged terms or concepts in e-commerce platforms to keep up-to-date, which is expensive and labor-intensive if it relies on manual maintenance and updates. Therefore, we target the taxonomy expansion task to attach new concepts to existing taxonomies automatically. In this paper, we present a self-supervised and user behavior-oriented product taxonomy expansion framework to append new concepts into existing taxonomies. Our framework extracts hyponymy relations that conform to users' intentions and cognition. Specifically, i) to fully exploit user behavioral information, we extract candidate hyponymy relations that match user interests from query-click concepts; ii) to enhance the semantic information of new concepts and better detect hyponymy relations, we model concepts and relations through both usergenerated content and structural information in existing taxonomies and user click logs, by leveraging Pre-trained Language Models and Graph Neural Network combined with Contrastive Learning; iii) to reduce the cost of dataset construction and overcome data skews, we construct a high-quality and balanced training dataset from existing taxonomy with no supervision. Extensive experiments on real-world product taxonomies in Meituan Platform, a leading Chinese vertical e-commerce platform to order take-out with more than 70 million daily active users, demonstrate the superiority of our proposed framework over state-of-the-art methods. Notably, our method enlarges the size of real-world product taxonomies from 39,263 to 94,698 relations with 88% precision. Our implementation is available: https://github.com/AdaCheng/Product Taxonomy Expansion.

GANTEE: Generative Adversarial Network for Taxonomy Enterance Evaluation

“…Open source (Yu et al 2020), user-clicked-logs (Cheng et al 2022 and parent-child information (Wang et al 2021(Wang et al , 2022 are all used by the previous methods to learn a better representation for concepts. Besides, there exist many studies use PLMs during the representation learning stage (Cheng et al 2022;Wang et al 2022;Takeoka, Akimoto, and Oyamada 2021), however, these methods only use the PLMs as a external information provider to learn a better representation.…”

“…However, we argue that these methods (Yu et al 2020;Wang et al 2022;Cheng et al 2022;Shen et al 2020;Mao et al 2020;Wang et al 2021;Zhang et al 2021) have two drawbacks when being applied to real applications. First, these methods suffer from low effectiveness in real applications due to limited ability to represent the semantics of concepts.…”

“…First, these methods suffer from low effectiveness in real applications due to limited ability to represent the semantics of concepts. Only sampling from existing taxonomy provides limited data to train representations for zero-shot concepts, where thousands times queries of the number of concepts in existing taxonomy are all unseen to the models (Cheng et al 2022). Second, these methods suffer from low efficiency in real applications because most of the query concepts are noisy.…”

“…Second, these methods suffer from low efficiency in real applications because most of the query concepts are noisy. For instance, in a leading ordering takeouts online platform, only thousands of concepts out of billions of queries are finally added to their taxonomy (Cheng et al 2022). Previous methods ignore the noisy concepts and waste much time finding the anchor concept for noisy query concepts.…”