Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Usage-centric Take on Intent Understanding in E-Commerce

Wendi Zhou, Tianyi Li, Pavlos Vougiouklis, Mark Steedman, Jeff Z. Pan

TL;DR

This paper focuses on predicative user intents as “how a customer uses a product”, and poses intent understanding as a natural language reasoning task, independent of product ontologies, and identifies two weaknesses of FolkScope, the SOTA E-Commerce Intent Knowledge Graph: category-rigidity and property-ambiguity.

Abstract

Identifying and understanding user intents is a pivotal task for E-Commerce. Despite its essential role in product recommendation and business user profiling analysis, intent understanding has not been consistently defined or accurately benchmarked. In this paper, we focus on predicative user intents as "how a customer uses a product", and pose intent understanding as a natural language reasoning task, independent of product ontologies. We identify two weaknesses of FolkScope, the SOTA E-Commerce Intent Knowledge Graph: category-rigidity and property-ambiguity. They limit its ability to strongly align user intents with products having the most desirable property, and to recommend useful products across diverse categories. Following these observations, we introduce a Product Recovery Benchmark featuring a novel evaluation framework and an example dataset. We further validate the above FolkScope weaknesses on this benchmark. Our code and dataset are available at https://github.com/stayones/Usgae-Centric-Intent-Understanding.

A Usage-centric Take on Intent Understanding in E-Commerce

TL;DR

This paper focuses on predicative user intents as “how a customer uses a product”, and poses intent understanding as a natural language reasoning task, independent of product ontologies, and identifies two weaknesses of FolkScope, the SOTA E-Commerce Intent Knowledge Graph: category-rigidity and property-ambiguity.

Abstract

Identifying and understanding user intents is a pivotal task for E-Commerce. Despite its essential role in product recommendation and business user profiling analysis, intent understanding has not been consistently defined or accurately benchmarked. In this paper, we focus on predicative user intents as "how a customer uses a product", and pose intent understanding as a natural language reasoning task, independent of product ontologies. We identify two weaknesses of FolkScope, the SOTA E-Commerce Intent Knowledge Graph: category-rigidity and property-ambiguity. They limit its ability to strongly align user intents with products having the most desirable property, and to recommend useful products across diverse categories. Following these observations, we introduce a Product Recovery Benchmark featuring a novel evaluation framework and an example dataset. We further validate the above FolkScope weaknesses on this benchmark. Our code and dataset are available at https://github.com/stayones/Usgae-Centric-Intent-Understanding.
Paper Structure (29 sections, 2 equations, 3 figures, 6 tables)

This paper contains 29 sections, 2 equations, 3 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Usage-Centric Intent Understanding
  3. FolkScope Analysis
  4. KG Refactoring
  5. Statistical Analysis
  6. Property-Ambiguity
  7. Category-Rigidity
  8. The Product Recovery Benchmark
  9. Benchmark Design
  10. Evaluation metric
  11. Experiments and Results
  12. Property-Ambiguity
  13. Category-Rigidity
  14. LLM Rerank
  15. Discussions and Conclusion
  16. ...and 14 more sections

Figures (3)

  • Figure 1: A graphic illustration of the usage-centric paradigm of intent understanding.
  • Figure 2: Histograms of Jensen-Shannon Divergence for each intent-category pair. Values are packed around 0: property-distributions of edge weights conditioned on intents are close to unconditioned frequency priors.
  • Figure 3: Histograms of category-entropy for each user intent. Values are concentrated at 0.0 and 0.7, meaning the intent is associated with only 1 / 2 categories.