System generates recommendation dialogues from search data and knowledge graphs
Dialogue Generation for Conversational Recommendation Using User Search Interactions and Knowledge Graphs
Artificial IntelligenceComputation and LanguageMachine Learning
Summary
Finding good recommendations in conversations is hard and usually requires experts to create example chats manually. The authors introduce a method called SearCl that automatically creates recommendation dialogues by looking at how users search and click on items online. Their system uses a state machine to guide the conversation flow and two AI agents that take turns playing the roles of seeker and recommender. By using knowledge graphs about the items being recommended, their dialogues are more informative and scored better than existing dialogue collections, all done without human annotation.
What this means in practice
- •For conversational ai developers: Generate realistic recommendation dialogues automatically using user search logs and item knowledge graphs to train dialogue systems without manual data labeling.
- •For e-commerce platform teams: Create knowledge-grounded conversational recommendation datasets that improve AI assistants’ ability to suggest products based on real user interactions and structured product data.$Commercial implications: Enables building commercial AI chatbots that provide personalized product suggestions by using automatically generated training dialogues reflecting real user behavior.
Authors
Minsung Hwang, Donggyu Yoon, Gowoon Hwang, Kwanghee Jang, Joyce Jiyoung Whang
Abstract
Large-scale conversational recommendation datasets typically require manual dialogue construction or annotation. We propose SearCl, which automatically generates conversational recommendation dialogues grounded in user search and click behaviors. SearCl segments user interactions into query sessions and derives each recommendation target from content selected after a search query. A configurable probabilistic state machine controls the dialogue flow, while two LLM agents acting as the seeker and recommender generate utterances according to turn-level constraints. To support recommendation, the recommender is provided with knowledge graphs constructed from metadata about the content items used in each dialogue, capturing their attributes and relationships. Using SearCl, we generate 3,538 dialogues from content search and click interaction data, covering 1,731 unique target items. Under an LLM-as-a-judge protocol, the resulting dialogues achieve higher quality scores than dialogues from existing benchmarks. These results demonstrate that SearCl enables controllable and knowledge-grounded dataset construction without human annotations.