Implicit Feedback: Techniques for Deployment and Evaluation

Опубликовано 6 сентября 2016, 6:06

Searchers can find the construction of query statements for submission to search systems a problematic activity. Implicit feedback models can proactively support searchers by passively observing interaction behavior and making recommendations about new query words to add, or retrieval strategies to adopt. Implicit feedback is typically gathered through monitoring behaviors such as bookmarking, saving or printing. Using document retention in this way can be worthwhile, but it is seldom observed in studies of Web search behavior and is highly context dependent. In this talk I will describe my approach to implicit feedback. I will give an overview of content-rich interfaces I have developed to improve the quality (and quantity) of searcher interaction, heuristic and probabilistic implicit feedback frameworks that use their interaction, and decision measures that approximate changes in searcher interests. I will describe how I evaluate my techniques with human subjects and simulations of searcher behavior. As an addendum, I will describe how this research fits into my other current work on exploratory search, story building for relevance feedback, query expansion, and positive and negative implicit feedback using eye tracking.