使用描述性图来提高对话推荐中的物品和情境理解（英文版）

英文标题：Improving Items and Contexts Understanding with Descriptive Graph for Conversational Recommendation中文摘要：本文介绍了一种新的会话式推荐系统框架 KLEVER，该框架通过构建物品描述图并联合学习物品和其相关上下文词语的嵌入来增强推荐和对话生成模块，从而提高推荐系统的性能。实验结果显示，KLEVER 在用户输入信息缺乏时表现出优异的性能。英文摘要：State-of-the-art methods on conversational recommender systems (CRS) leverageexternal knowledge to enhance both items' and contextual words' representationsto achieve high quality recommendations and responses generation. However, therepresentations of the items and wor