Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Intention Knowledge Graph Construction for User Intention Relation Modeling

Jiaxin Bai, Zhaobo Wang, Junfei Cheng, Dan Yu, Zerui Huang, Weiqi Wang, Xin Liu, Chen Luo, Yanming Zhu, Bo Li, Yangqiu Song

TL;DR

This paper introduces a framework to automatically generate an intention knowledge graph, capturing connections between user intentions, using the Amazon m2 dataset, and constructs an intention graph with 351 million edges, demonstrating high plausibility and acceptance.

Abstract

Understanding user intentions is challenging for online platforms. Recent work on intention knowledge graphs addresses this but often lacks focus on connecting intentions, which is crucial for modeling user behavior and predicting future actions. This paper introduces a framework to automatically generate an intention knowledge graph, capturing connections between user intentions. Using the Amazon m2 dataset, we construct an intention graph with 351 million edges, demonstrating high plausibility and acceptance. Our model effectively predicts new session intentions and enhances product recommendations, outperforming previous state-of-the-art methods and showcasing the approach's practical utility.

Intention Knowledge Graph Construction for User Intention Relation Modeling

TL;DR

This paper introduces a framework to automatically generate an intention knowledge graph, capturing connections between user intentions, using the Amazon m2 dataset, and constructs an intention graph with 351 million edges, demonstrating high plausibility and acceptance.

Abstract

Understanding user intentions is challenging for online platforms. Recent work on intention knowledge graphs addresses this but often lacks focus on connecting intentions, which is crucial for modeling user behavior and predicting future actions. This paper introduces a framework to automatically generate an intention knowledge graph, capturing connections between user intentions. Using the Amazon m2 dataset, we construct an intention graph with 351 million edges, demonstrating high plausibility and acceptance. Our model effectively predicts new session intentions and enhances product recommendations, outperforming previous state-of-the-art methods and showcasing the approach's practical utility.

Paper Structure

This paper contains 42 sections, 1 equation, 7 figures, 18 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. IGC-RC Framework
  4. Intention Generation
  5. Intention Relation Classification
  6. Intention Conceptualization
  7. Intrinsic Evaluation
  8. Human Evaluation
  9. Intention Generation
  10. Intention Relation Classification
  11. Intention Conceptualization
  12. Automatic Evaluation
  13. Intention Prediction
  14. Conceptualization Prediction
  15. Product Recovery
  16. ...and 27 more sections

Figures (7)

  • Figure 1: The structure of our knowledge graph. Part (A) shows an example of user behaviors within a session. Our knowledge graph emphasizes building relations between different intentions. In Part (B), we establish commonsense relations, highlighting the temporality and causality of intentions. Part (C) focuses on using conceptualization to connect various intentions.
  • Figure 2: Ablation results of different variants. This demonstrates that our intention knowledge graph significantly enhances recommendation performance compared to SASRec. Both intention conceptualization and concept relations effectively improve results, with each type of relation contributing uniquely to different metrics. This highlights the importance of incorporating diverse nodes and relations in the knowledge graph.
  • Figure 3: This figure shows the prompts we use to make LLM understand and generate intentions from user sessions.
  • Figure 4: This figure shows our prompts to make LLM conceptualize the user intentions.
  • Figure 5: This figure shows an example annotation question for the quality of session intention generation.
  • ...and 2 more figures