The paper presents IAI MovieBot 2.0, an improved version of the conversational movie recommender system, emphasizing trainable neural components, transparent user modeling, and enhancements in the user interface and research infrastructure. The authors address the limitations of existing open-source conversational recommender systems by introducing new features to facilitate user-facing experiments and personalized recommendations.
The enhancements include new natural language understanding and dialogue manager components trained using deep learning approaches, a user model for storing long-term preferences, a new web widget for multimodal interactions, and an updated codebase utilizing DialogueKit2 library. These improvements aim to make IAI MovieBot 2.0 more modular, adaptable, and user-friendly.
The paper also discusses related work in conversational recommender systems, highlighting the scarcity of operational systems suitable for comprehensive studies and comparing IAI MovieBot with other research prototypes like Vote Goat and DAGFiNN. Additionally, it evaluates the performance of the newly added neural components through experiments on natural language understanding and dialogue policy learning.
Egy másik nyelvre
a forrásanyagból
arxiv.org
Mélyebb kérdések