Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

IAI MovieBot 2.0: An Enhanced Research Platform with Trainable Neural Components and Transparent User Modeling

Nolwenn Bernard, Ivica Kostric, Krisztian Balog

TL;DR

The paper tackles the lack of open, end-to-end research platforms for conversational recommender systems by introducing IAI MovieBot 2.0, which integrates trainable neural NLU and dialogue policy, a transparent persistent user model, and improved UI and infrastructure. It deploys JointBERT with a CRF layer for unified intent-slot handling and RL-based policy learning (A2C and DQN) within an agenda-based user simulator, highlighting both potential and current limitations of neural components. The work demonstrates the platform's experimental capabilities and emphasizes modularity through DialogueKit and REST/socket.io deployment to enable reproducible, user-facing studies, while acknowledging that neural NLU may require more data and refined state representations to outperform handcrafted rules. Overall, IAI MovieBot 2.0 provides a scalable, extensible foundation for investigating CRSs in research settings and paves the way for future enhancements such as neural NLG and richer recommendation components.

Abstract

While interest in conversational recommender systems has been on the rise, operational systems suitable for serving as research platforms for comprehensive studies are currently lacking. This paper introduces an enhanced version of the IAI MovieBot conversational movie recommender system, aiming to evolve it into a robust and adaptable platform for conducting user-facing experiments. The key highlights of this enhancement include the addition of trainable neural components for natural language understanding and dialogue policy, transparent and explainable modeling of user preferences, along with improvements in the user interface and research infrastructure.

IAI MovieBot 2.0: An Enhanced Research Platform with Trainable Neural Components and Transparent User Modeling

TL;DR

The paper tackles the lack of open, end-to-end research platforms for conversational recommender systems by introducing IAI MovieBot 2.0, which integrates trainable neural NLU and dialogue policy, a transparent persistent user model, and improved UI and infrastructure. It deploys JointBERT with a CRF layer for unified intent-slot handling and RL-based policy learning (A2C and DQN) within an agenda-based user simulator, highlighting both potential and current limitations of neural components. The work demonstrates the platform's experimental capabilities and emphasizes modularity through DialogueKit and REST/socket.io deployment to enable reproducible, user-facing studies, while acknowledging that neural NLU may require more data and refined state representations to outperform handcrafted rules. Overall, IAI MovieBot 2.0 provides a scalable, extensible foundation for investigating CRSs in research settings and paves the way for future enhancements such as neural NLG and richer recommendation components.

Abstract

While interest in conversational recommender systems has been on the rise, operational systems suitable for serving as research platforms for comprehensive studies are currently lacking. This paper introduces an enhanced version of the IAI MovieBot conversational movie recommender system, aiming to evolve it into a robust and adaptable platform for conducting user-facing experiments. The key highlights of this enhancement include the addition of trainable neural components for natural language understanding and dialogue policy, transparent and explainable modeling of user preferences, along with improvements in the user interface and research infrastructure.
Paper Structure (13 sections, 1 figure, 2 tables)

This paper contains 13 sections, 1 figure, 2 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. The Existing IAI MovieBot
  4. IAI MovieBot 2.0 Extensions
  5. Neural Natural Language Understanding
  6. Dialogue Policy Learning
  7. User Modeling
  8. User Interface and Experience
  9. Research Infrastructure Improvements
  10. Experiments
  11. Natural Language Understanding
  12. Dialogue Policy Learning
  13. Conclusion

Figures (1)

  • Figure 1: Overview of IAI MovieBot 2.0 architecture. Blue components are inherited from Habib:2020:CIKM and the green ones are new additions. Training utilities are available for components marked with a star (*).