MobileConvRec: A Conversational Dataset for Mobile Apps Recommendations

现有的推荐系统主要分为两种范式：1-基于历史用户-物品交互的推荐和2-对话式推荐。对话式推荐系统可以在用户和系统之间进行自然语言对话，使系统能够征求用户的明确需求，同时使用户可以查询推荐并提供反馈。由于自然语言处理方面的重大进展，对话式推荐系统已经受到了重视。现有的对话式推荐数据集极大地促进了它们各自领域的研究。尽管近年来移动用户和应用程序呈指数增长，但对话式移动应用程序推荐系统的研究面临着重大的限制。这种限制主要归因于缺乏专门针对移动应用程序的高质量基准数据集。为了促进对话式移动应用程序推荐的研究，我们介绍了MobileConvRec。MobileConvRec通过利用在Google Play商店上的移动应用程序的真实用户交互，最初在大规模移动应用程序推荐数据集MobileRec中捕获的数据，来模拟对话。所提出的对话式推荐数据集通过将顺序用户-物品交互（反映隐含用户偏好）与全面的多轮对话相结合，有效地把握了明确的用户需求。MobileConvRec包括超过12K个跨45个应用类别的多轮推荐相关对话。此外，MobileConvRec还为每个应用程序提供了丰富的元数据，例如权限数据、安全和隐私相关信息以及应用程序的二进制可执行文件等。我们通过对几个预训练的大型语言模型的比较研究，证明了MobileConvRec可以作为对话式移动应用程序推荐的优秀测试平台。

MobileConvRec: A Multi-Turn Conversational Recommendation Dataset for Mobile Apps

MobileConvRec is a new conversational recommendation dataset specifically tailored for mobile apps, which synergizes implicit and explicit user preferences to effectively grasp user needs.

MobileConvRec simulates conversations by leveraging real user interactions with mobile apps on the Google Play store, and consists of over 12K multi-turn recommendation-related conversations spanning 45 app categories. The dataset presents rich metadata for each app such as permissions data, security and privacy-related information, and binary executables of apps. MobileConvRec can serve as an excellent testbed for conversational mobile app recommendation through a comparative study of several pre-trained large language models.

Existing recommendation systems have focused on historical user-item interaction-based recommendations and conversational recommendations. Recent research in conversational recommendation systems has gained prominence due to advancements in natural language processing. However, research in conversational mobile app recommender systems has faced substantial constraints due to the lack of high-quality benchmark datasets specifically tailored for mobile apps.