A5057182567- Multimodal Machine Learning Applications
- Recommender Systems and Techniques
- Topic Modeling
Korea Telecom (South Korea)
2023
Conversational Recommender Systems (CRS) aim to provide tailored recommendation responses via a chat interface, including both the user's preferred item and its accompanying explanation. However, due generative nature, CRS are prone responding with factually incorrect explanations (i.e., hallucinations). To solve this problem, we propose incorporating passage retrieval module into objective of enhancing factuality informativeness system responses. Specifically, outline essential directions...
For high-quality conversational recommender systems (CRS), it is important to recommend the suitable items by capturing items' features mentioned in dialog and explain appropriate ones among various of recommended item. We argue that CRS model should be a domain-expert who (1) knowledgeable about relationships between their (2) able item with its relevant context. To this end, we propose novel framework, named as LATTE, pre-train each core module (i.e., recommendation conversation module)...