Bridging Conversational and Collaborative Signals for Conversational Recommendation
Fuente:
arXiv
Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866909487551479808 |
|---|---|
| author | Rabiah, Ahmad Bin Sadeq, Nafis McAuley, Julian |
| author_facet | Rabiah, Ahmad Bin Sadeq, Nafis McAuley, Julian |
| contents | Conversational recommendation systems (CRS) leverage contextual information from conversations to generate recommendations but often struggle due to a lack of collaborative filtering (CF) signals, which capture user-item interaction patterns essential for accurate recommendations. We introduce Reddit-ML32M, a dataset that links Reddit conversations with interactions on MovieLens 32M, to enrich item representations by leveraging collaborative knowledge and addressing interaction sparsity in conversational datasets. We propose an LLM-based framework that uses Reddit-ML32M to align LLM-generated recommendations with CF embeddings, refining rankings for better performance. We evaluate our framework against three sets of baselines: CF-based recommenders using only interactions from CRS tasks, traditional CRS models, and LLM-based methods relying on conversational context without item representations. Our approach achieves consistent improvements, including a 12.32% increase in Hit Rate and a 9.9% improvement in NDCG, outperforming the best-performing baseline that relies on conversational context but lacks collaborative item representations. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2412_06949 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Bridging Conversational and Collaborative Signals for Conversational Recommendation Rabiah, Ahmad Bin Sadeq, Nafis McAuley, Julian Information Retrieval Artificial Intelligence Machine Learning Conversational recommendation systems (CRS) leverage contextual information from conversations to generate recommendations but often struggle due to a lack of collaborative filtering (CF) signals, which capture user-item interaction patterns essential for accurate recommendations. We introduce Reddit-ML32M, a dataset that links Reddit conversations with interactions on MovieLens 32M, to enrich item representations by leveraging collaborative knowledge and addressing interaction sparsity in conversational datasets. We propose an LLM-based framework that uses Reddit-ML32M to align LLM-generated recommendations with CF embeddings, refining rankings for better performance. We evaluate our framework against three sets of baselines: CF-based recommenders using only interactions from CRS tasks, traditional CRS models, and LLM-based methods relying on conversational context without item representations. Our approach achieves consistent improvements, including a 12.32% increase in Hit Rate and a 9.9% improvement in NDCG, outperforming the best-performing baseline that relies on conversational context but lacks collaborative item representations. |
| title | Bridging Conversational and Collaborative Signals for Conversational Recommendation |
| topic | Information Retrieval Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2412.06949 |