Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Large Language Models for Intent-Driven Session Recommendations

Zhu Sun, Hongyang Liu, Xinghua Qu, Kaidong Feng, Yan Wang, Yew-Soon Ong

TL;DR

This work tackles intent-aware session recommendation by addressing the fixed, uniform-intent assumption and opaque latent representations. It introduces PO4ISR, a prompt-optimization framework that uses PromptInit to seed semantic intent understanding, PromptOpt for iterative self-reflection and refinement, and PromptSel for cross-domain prompt selection, guided by LLMs. Across three real-world datasets, PO4ISR significantly outperforms baselines (average HR and NDCG improvements of roughly 57% and 61%, respectively) and demonstrates strong cross-domain generalization, interpretability, and robustness, albeit with some hallucination risks inherent to LLMs. The study highlights the importance of prompt quality and iterative optimization for unlocking LLMs’ reasoning in ISR, and points to future work on reducing hallucinations and further improving cross-domain prompt strategies.

Abstract

Intent-aware session recommendation (ISR) is pivotal in discerning user intents within sessions for precise predictions. Traditional approaches, however, face limitations due to their presumption of a uniform number of intents across all sessions. This assumption overlooks the dynamic nature of user sessions, where the number and type of intentions can significantly vary. In addition, these methods typically operate in latent spaces, thus hinder the model's transparency.Addressing these challenges, we introduce a novel ISR approach, utilizing the advanced reasoning capabilities of large language models (LLMs). First, this approach begins by generating an initial prompt that guides LLMs to predict the next item in a session, based on the varied intents manifested in user sessions. Then, to refine this process, we introduce an innovative prompt optimization mechanism that iteratively self-reflects and adjusts prompts. Furthermore, our prompt selection module, built upon the LLMs' broad adaptability, swiftly selects the most optimized prompts across diverse domains. This new paradigm empowers LLMs to discern diverse user intents at a semantic level, leading to more accurate and interpretable session recommendations. Our extensive experiments on three real-world datasets demonstrate the effectiveness of our method, marking a significant advancement in ISR systems.

Large Language Models for Intent-Driven Session Recommendations

TL;DR

This work tackles intent-aware session recommendation by addressing the fixed, uniform-intent assumption and opaque latent representations. It introduces PO4ISR, a prompt-optimization framework that uses PromptInit to seed semantic intent understanding, PromptOpt for iterative self-reflection and refinement, and PromptSel for cross-domain prompt selection, guided by LLMs. Across three real-world datasets, PO4ISR significantly outperforms baselines (average HR and NDCG improvements of roughly 57% and 61%, respectively) and demonstrates strong cross-domain generalization, interpretability, and robustness, albeit with some hallucination risks inherent to LLMs. The study highlights the importance of prompt quality and iterative optimization for unlocking LLMs’ reasoning in ISR, and points to future work on reducing hallucinations and further improving cross-domain prompt strategies.

Abstract

Intent-aware session recommendation (ISR) is pivotal in discerning user intents within sessions for precise predictions. Traditional approaches, however, face limitations due to their presumption of a uniform number of intents across all sessions. This assumption overlooks the dynamic nature of user sessions, where the number and type of intentions can significantly vary. In addition, these methods typically operate in latent spaces, thus hinder the model's transparency.Addressing these challenges, we introduce a novel ISR approach, utilizing the advanced reasoning capabilities of large language models (LLMs). First, this approach begins by generating an initial prompt that guides LLMs to predict the next item in a session, based on the varied intents manifested in user sessions. Then, to refine this process, we introduce an innovative prompt optimization mechanism that iteratively self-reflects and adjusts prompts. Furthermore, our prompt selection module, built upon the LLMs' broad adaptability, swiftly selects the most optimized prompts across diverse domains. This new paradigm empowers LLMs to discern diverse user intents at a semantic level, leading to more accurate and interpretable session recommendations. Our extensive experiments on three real-world datasets demonstrate the effectiveness of our method, marking a significant advancement in ISR systems.
Paper Structure (22 sections, 6 figures, 6 tables, 2 algorithms)

This paper contains 22 sections, 6 figures, 6 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Session-based Recommendation
  4. Intent-aware Session Recommendation
  5. LLM-based Session Recommendation
  6. The Proposed Method
  7. Prompt Initialization (PromptInit)
  8. Prompt Optimization (PromptOpt)
  9. Prompt Selection (PromptSel)
  10. Experiments and Results
  11. Experimental Setup
  12. Datasets
  13. Baselines
  14. Parameter Settings
  15. Evaluation Metrics
  16. ...and 7 more sections

Figures (6)

  • Figure 1: Examples of user sessions with various intents.
  • Figure 2: The overall architecture of our proposed PO4ISR paradigm.
  • Figure 3: Performance of Top-1&-2 prompts (same domain).
  • Figure 4: Performance of Top-1 prompt (cross-domain).
  • Figure 5: Performance of Top-1 prompt from each domain in the cross-domain scenario on the test set (seed 0).
  • ...and 1 more figures