Reproduce “COLA: Improving Conversational Recommender Systems by Collaborative Augmentation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Dongding Lin, Jian Wang, Wenjie Li
COLA introduces a collaborative augmentation method that enhances item and user representations in conversational recommender systems by incorporating global popularity signals and insights from similar historical conversations. It achieves state-of-the-art performance on the REDIAL and TG-ReDial datasets, improving recommendation recall (e.g., 4.5% in R@10 on REDIAL) and generating more fluent and diverse responses.
Key insights
- Integrating global item popularity signals, extracted from an interactive user-item graph derived from the entire conversation corpus, significantly enhances item representations.
- Augmenting user preferences by retrieving and incorporating information from similar historical conversations effectively addresses the sparsity of information available in single dialogue turns.
- Adapting collaborative filtering principles to the conversational domain by constructing implicit user-item interaction graphs improves both item and user modeling in Conversational Recommender Systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.