Building Realistic Simulations for Interactive Information Retrieval

Simulation has been used within the field of Information Retrieval (IR) for many years to evaluate retrieval models and other aspects of the wider IR process. In recent years, there has been a renewed interest towards using simulation for Interactive Information Retrieval (IIR), an area which focuses on the study of human interactions with IR systems. A variety of different interaction models (e.g. click models) associated with behavioural aspects of searchers have over time been developed and evaluated using simulation in order for us to better understand the complex processes involved. Despite these advances, such models are still relatively naïve, and further work is required to make simulations of searchers more realistic. To this end, this project seeks to build more realistic simulations, using a more Complex Searcher Model (CSM). Within the CSM, each component and decision point can be varied and customised as required. The CSM can then be instantiated using components that are grounded from empirical evidence based upon actual real-world searcher behaviour and interaction data