Search Agent Model: A Conceptual Framework for Search by Algorithms and Agent Systems.
Jeff Dalton, John F. Foley · ENLIGHTEN (Jurnal Bimbingan dan Konseling Islam) · 2018
We describe challenges using search systems designed for algorithms and agent systems rather than humans.As information access systems become more complex the users of retrieval systems are increasingly shifting from humans to agents that use search as a sensor for acquiring and interpreting knowledge of the world.First, we discuss work on prior search applications that fit into this agent model of search.We identify key challenges for current and future search systems including: confidence estimation, task state, and expressing complex long-term retrieval models.We propose a conceptual framework for understanding and creating search-focused agents that addresses these challenges, the Search Agent Model (SAM).SAM provides a shared model for complex search tasks requiring a variety of information processing and orchestration.Its main components include formalizing task state, an action policy, and a query language for agent interaction.We describe how this proposed agent model provides a roadmap for future research and system design in search.