Recommendation systems have evolved to offer more comprehensive, topic-focused services, with conversational recommendation systems (CRS) advancing from simple retrieval-based methods to sophisticated agentic systems featuring advanced reasoning and self-correction. However, these agentic systems often suffer from high response latency, which poses a challenge for real-time conversational interactions. AdaptJobRec addresses this issue by introducing a user query complexity identification mechanism that optimizes the system’s response strategy based on query difficulty. For simple queries, the system quickly selects the appropriate recommendation tool to provide rapid responses, minimizing latency. For more complex queries, AdaptJobRec employs a memory processing module to filter relevant chat history, followed by an intelligent task decomposition planner that breaks down the query into manageable tasks executed by personalized recommendation tools. This multi-stage approach allows the system to maintain high accuracy without sacrificing speed. The system was evaluated using Walmart’s real-world career recommendation scenarios, demonstrating a substantial reduction in average response latency by up to 53.3% compared to existing baselines. Additionally, it significantly improved the accuracy of job recommendations, highlighting its practical effectiveness. AdaptJobRec’s design offers a promising balance between handling complex conversational queries and delivering timely, personalized recommendations. This advancement has implications for enhancing user experience in conversational AI systems, particularly in career guidance and other recommendation domains. Future work could explore further optimization of the agentic system and expansion to other recommendation contexts.
👉 Pročitaj original: arXiv AI Papers
