Structured text representations and relevance judgments: an exploratory study.

Abstract

This paper presents results of an exploratory study on using a structured text and query model for analyzing relevance judgments made by users on retrieved texts The structured model moves beyond keyword based text representation and retrieval. In particular it allows texts and queries to be represented by concepts embedded within the structural framework of complex objects. These objects preserves important text-based links between the selected concepts. The current study reveals a tendency for user judged relevance potential to decrease as the structure-based similarity between query and text decreases. This indicates that the structure-based retrieval model has the potential to conduct more precise retrieval than keyword based models. This supports our overall research goal of developing and investigating the structured retrieval model.