Enhancing User Intent for Recommendation Systems via Large Language Models

TL;DR

This work tackles the challenge of dynamic user intent in recommendations by proposing DUIP, a framework that fuses LSTM-based intent modeling with an LLM-driven next-item predictor. The LSTM processes sequential interactions to produce a hidden state $h_t$, which seeds a soft prompt $P=f(h_t)$ that, together with hard prompts, conditions a Large Language Model to predict the next item via $\hat{y}=\arg\max_{y\in Y} P(y|P)$. Empirically, DUIP outperforms a broad set of baselines on ML-1M, Games, and Bundle datasets, addressing cold-start and real-time adaptation through dynamic prompting. The results suggest that dynamic prompts guided by evolving user intent can yield more accurate, context-aware recommendations and motivate further work on cross-modal data, cross-domain transfer, and scalable online deployment.

Abstract

Recommendation systems play a critical role in enhancing user experience and engagement in various online platforms. Traditional methods, such as Collaborative Filtering (CF) and Content-Based Filtering (CBF), rely heavily on past user interactions or item features. However, these models often fail to capture the dynamic and evolving nature of user preferences. To address these limitations, we propose DUIP (Dynamic User Intent Prediction), a novel framework that combines LSTM networks with Large Language Models (LLMs) to dynamically capture user intent and generate personalized item recommendations. The LSTM component models the sequential and temporal dependencies of user behavior, while the LLM utilizes the LSTM-generated prompts to predict the next item of interest. Experimental results on three diverse datasets ML-1M, Games, and Bundle show that DUIP outperforms a wide range of baseline models, demonstrating its ability to handle the cold-start problem and real-time intent adaptation. The integration of dynamic prompts based on recent user interactions allows DUIP to provide more accurate, context-aware, and personalized recommendations. Our findings suggest that DUIP is a promising approach for next-generation recommendation systems, with potential for further improvements in cross-modal recommendations and scalability.