# Pre-train, Prompt and Recommendation: A Comprehensive Survey of Language Modelling Paradigm Adaptations in Recommender Systems

**Authors:** Peng Liu, Lemei Zhang, Jon Atle Gulla  
**arXiv:** [2302.03735](https://arxiv.org/abs/2302.03735)  
**Date:** 2023-02-07  
**Summary page:** https://www.summarizepaper.com/en/paper/2302.03735v2/

## Key points

- Pre-trained Language Models (PLMs) have revolutionized Natural Language Processing (NLP)
- PLMs have been extended to the recommendation domain and are considered a promising approach
- "Pre-train, Prompt and Recommendation: A Comprehensive Survey of Language Modelling Paradigm Adaptations in Recommender Systems" investigates how to extract and transfer knowledge from pre-trained models learned by different PLM-related training paradigms to improve recommendation performance
- The authors propose an orthogonal taxonomy to divide existing PLM-based recommender systems with respect to their training strategies and objectives
- The paper analyzes and summarizes the connection between PLM-based training paradigms and different input data types for recommender systems
- The paper elaborates on open issues and future research directions in this field
- This comprehensive survey provides insights into how pre-trained language models can be adapted for use in recommender systems, improving recommendation performance across various domains while addressing challenges such as sparsity, efficiency, and effectiveness
- The proposed taxonomy offers a framework for understanding the different approaches used in PLM-based recommender systems while highlighting areas where further research is needed.

## Summary

The emergence of Pre-trained Language Models (PLMs) has revolutionized the field of Natural Language Processing (NLP) by learning universal representations on large corpora in a self-supervised manner. This approach has been extended to the recommendation domain and is considered a promising approach by both academia and industry. In their paper, "Pre-train, Prompt and Recommendation: A Comprehensive Survey of Language Modelling Paradigm Adaptations in Recommender Systems," Peng Liu, Lemei Zhang, and Jon Atle Gulla systematically investigate how to extract and transfer knowledge from pre-trained models learned by different PLM-related training paradigms to improve recommendation performance from various perspectives such as generality, sparsity, efficiency, and effectiveness. The authors propose an orthogonal taxonomy to divide existing PLM-based recommender systems with respect to their training strategies and objectives. They analyze and summarize the connection between PLM-based training paradigms and different input data types for recommender systems. The paper also elaborates on open issues and future research directions in this vibrant field. This comprehensive survey provides insights into how pre-trained language models can be adapted for use in recommender systems. By leveraging the power of these models, researchers can improve recommendation performance across various domains while addressing challenges such as sparsity, efficiency, and effectiveness. The proposed taxonomy offers a framework for understanding the different approaches used in PLM-based recommender systems while highlighting areas where further research is needed. Overall, this paper offers valuable insights into the potential of pre-trained language models for improving recommendation systems.