Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Reformulating Sequential Recommendation: Learning Dynamic User Interest with Content-enriched Language Modeling

Junzhe Jiang, Shang Qu, Mingyue Cheng, Qi Liu, Zhiding Liu, Hao Zhang, Rujiao Zhang, Kai Zhang, Rui Li, Jiatong Li, Min Gao

TL;DR

This work reframes sequential recommendation as a text-generation task by integrating pretrained language models with content-enriched prompts. It introduces knowledge prompts to inject domain knowledge and reasoning prompts to fuse domain content with user history, while an item-mapping step translates generated text into concrete items. Empirical results on MovieLens, MIND, and Goodreads show that LANCER achieves state-of-the-art performance in top-k recall and ranking metrics, with ablations confirming the contributions of both prompt types and decoding strategies. The approach demonstrates the potential of bridging language-model capabilities and recommender systems, enabling dynamic, content-aware user understanding and personalized recommendations with implications for more human-like, context-rich suggestions.

Abstract

Recommender systems are indispensable in the realm of online applications, and sequential recommendation has enjoyed considerable prevalence due to its capacity to encapsulate the dynamic shifts in user interests. However, previous sequential modeling methods still have limitations in capturing contextual information. The primary reason is the lack of understanding of domain-specific knowledge and item-related textual content. Fortunately, the emergence of powerful language models has unlocked the potential to incorporate extensive world knowledge into recommendation algorithms, enabling them to go beyond simple item attributes and truly understand the world surrounding user preferences. To achieve this, we propose LANCER, which leverages the semantic understanding capabilities of pre-trained language models to generate personalized recommendations. Our approach bridges the gap between language models and recommender systems, resulting in more human-like recommendations. We demonstrate the effectiveness of our approach through a series of experiments conducted on multiple benchmark datasets, showing promising results and providing valuable insights into the influence of our model on sequential recommendation tasks. Furthermore, our experimental codes are publicly available at https://github.com/Gnimixy/lancer.

Reformulating Sequential Recommendation: Learning Dynamic User Interest with Content-enriched Language Modeling

TL;DR

This work reframes sequential recommendation as a text-generation task by integrating pretrained language models with content-enriched prompts. It introduces knowledge prompts to inject domain knowledge and reasoning prompts to fuse domain content with user history, while an item-mapping step translates generated text into concrete items. Empirical results on MovieLens, MIND, and Goodreads show that LANCER achieves state-of-the-art performance in top-k recall and ranking metrics, with ablations confirming the contributions of both prompt types and decoding strategies. The approach demonstrates the potential of bridging language-model capabilities and recommender systems, enabling dynamic, content-aware user understanding and personalized recommendations with implications for more human-like, context-rich suggestions.

Abstract

Recommender systems are indispensable in the realm of online applications, and sequential recommendation has enjoyed considerable prevalence due to its capacity to encapsulate the dynamic shifts in user interests. However, previous sequential modeling methods still have limitations in capturing contextual information. The primary reason is the lack of understanding of domain-specific knowledge and item-related textual content. Fortunately, the emergence of powerful language models has unlocked the potential to incorporate extensive world knowledge into recommendation algorithms, enabling them to go beyond simple item attributes and truly understand the world surrounding user preferences. To achieve this, we propose LANCER, which leverages the semantic understanding capabilities of pre-trained language models to generate personalized recommendations. Our approach bridges the gap between language models and recommender systems, resulting in more human-like recommendations. We demonstrate the effectiveness of our approach through a series of experiments conducted on multiple benchmark datasets, showing promising results and providing valuable insights into the influence of our model on sequential recommendation tasks. Furthermore, our experimental codes are publicly available at https://github.com/Gnimixy/lancer.
Paper Structure (17 sections, 8 equations, 3 figures, 2 tables)

This paper contains 17 sections, 8 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Model Architecture.
  4. Knowledge Prompt.
  5. Reasoning Prompt.
  6. Item Mapping.
  7. Experiments
  8. Experimental Setup
  9. Datasets.
  10. Evaluation Metrics.
  11. Implementation Details.
  12. Compared Methods.
  13. Results and Analyses
  14. Recommendation Performance Evaluation.
  15. Case Study.
  16. ...and 2 more sections

Figures (3)

  • Figure 1: Comparison of sequence recommendation with and without content-enriched information. It can be seen that enriched content information provides more information about the content of the movies, which helps to reduce ambiguity and improve the accuracy of the recommendations.
  • Figure 2: Illustration of the proposed LANCER, where the dashed line indicates the trainable parameters and the solid line indicates the frozen parameters.
  • Figure 3: Case study of three users from the MovieLens dataset in terms of top-5 ranked predictions. The texts in green indicate predictions that map to the target.