InfoCrystal A Visual Tool For Information Retrieval Ph.D. Research and Thesis at MIT, 1995 Table of Contents | Papers | Slides | About
Abstract The InfoCrystal is a novel representation that uses a simple visual metaphor to help users deal with some of the complexities inherent in information retrieval. As a visualization tool, it can display all the possible binary as well as continuous relationships among N concepts. As a visual query language, the InfoCrystal enables users to formulate both Boolean and vector space queries graphically. Hence, it provides a visual framework that unifies the complementary Boolean and Partial Matching approaches and allows users to take advantage of their respective strengths. The InfoCrystal acts like a Boolean Calculator and users can use it to employ the expressive power of the Boolean retrieval approach and its broadening / narrowing techniques in a visual way. Further, users can assign relevance weights to the concepts and formulate weighted queries by interacting with a threshold slider. The InfoCrystal offers the added advantage that users can control in a visual way how to translate weighted queries into Boolean queries. Finally, arbitrarily complex queries can be created by using the InfoCrystals as building blocks and organizing them in a hierarchical structure. A user study tested a specific aspect of the InfoCrystal interface by comparing it with a standard Boolean retrieval interface. Although this study did not test all the valuable features of the InfoCrystal, it produced the following useful results: 1) It showed that novice users, who received only a short tutorial, could successfully use the novel InfoCrystal interface. 2) The study showed that the InfoCrystal, even at an early stage of development, performed as well as the familiar Boolean interface, although the study was biased in favor of the Boolean mode. 3) The user feedback concerning the InfoCrystal interface was very encouraging and it helped to pinpoint possible improvements. The InfoCrystal has broad applications because it offers a "visual machinery" to compare and relate any number of ordinary or fuzzy sets of arbitrary data items. It opens up new possibilities for complex data explorations. The InfoCrystal enables users to integrate and explore information retrieved by different methods or from different sources in a flexible, dynamic and interactive way. This thesis describes research done at the Massachusetts Institute of Technology within the Center for Educational Computing Initiatives. Support for the research described in this thesis has been generously provided by the UBILAB of the Union Bank of Switzerland.
Table of Contents 1 Introduction ... 1 1.1 Information Visualization ... 2 1.2 Information Retrieval ... 4 1.3 Goal of the Thesis ... 5 1.4 Thesis Organization ... 6 1.5 Concrete Example ... 6 2 Information Retrieval Models ... 9 2.1 Introduction ... 9 2.2 General Model of Information Retrieval ... 9 2.3 Major Information Retrieval Models ... 13 2.3.1 Boolean Retrieval ... 13 2.3.1.1 Standard Boolean ... 14 2.3.1.2 Narrowing and Broadening Techniques ... 16 2.3.1.3 Conclusion ... 30 3 InfoCrystal ... 33 3.1 Introduction ... 33 3.2 2D versus 3D Visualization ... 33 3.3 Visualizing Relationships ... 34 3.4 Rank Layout Algorithm ...37 3.5 Example Revisited ...49 3.6 The Design Process of the InfoCrystal ... 51 3.6.1 The First Designs for the InfoCrystal ... 51 3.6.2 InfoCrystal Networks ... 53 3.6.3 Combining the InfoCrystal with Venn Diagrams ... 54 4 Visualizing Boolean Queries ... 59 4.1 Introduction ... 59 4.2 Query Space Visualized by the InfoCrystal ... 61 4.2.1 Ways of Specifying a Boolean Query ... 62 4.2.2 InfoCrystal as a Boolean Calculator ... 62 4.3 Creating Complex and Nested Queries ... 64 4.4 The Outliner Tool ... 70 4.5 Narrowing and Broadening Techniques ... 71 5 Visualizing Weighted Queries ... 77 5.1 Introduction ... 77 5.2 Formulating Weighted Queries using the InfoCrystal ... 78 5.3 The Bull's...Eye Layout ... 82 5.4 The Expressive Limits of Weighted Queries ... 85 5.5 Possible Alternative for Specifying Weighted Queries ... 87 5.6 Discussion ... 88 6 Visualizing Vector Space Queries ... 89 6.1 Introduction ... 89 6.2 Visualizing Any Ranking Function ... 91 6.3 The Continuous Bull's...Eye Mapping ... 96 6.4 Discussion ... 99 7 InfoCrystal Software ... 101 7.1 Introduction ... 101 7.2 InfoCrystal Software in Pictures ... 104 7.3 How to Drop an Input from an InfoCrystal ? ... 118 7.4 How to Add an Input to an InfoCrystal ? ... 118 7.5 How to Update the Selection Pattern in a Modified InfoCrystal ? ... 119 8 Experimental Evaluation ... 121 8.1 Introduction ... 121 8.2 Experimental Design ... 122 8.3 Experimental Analysis ... 131 8.3.1 Paired...Difference T...Test ... 133 8.3.2 Analysis of Variance ... 133 8.4 Analysis of the Experimental Results ... 137 8.4.1 Results for the Recognition Task ... 137 8.4.1.1 Categorical Paired...Difference Scores ... 138 8.4.1.2 Time Measurements ... 140 8.4.1.3 Discussion ... 142 8.4.2 Results for the Generation Task ... 145 8.4.2.1 Generation Task Biased in Favor of Boolean Query Language ... 145 8.4.2.2 Categorical Paired...Difference Scores ... 147 8.4.2.3 Continuous Paired...Difference Scores ... 150 8.4.2.4 Time Measurements ... 152 8.4.2.5 Discussion ... 155 8.5 Lessons Learned and General Discussion ... 158 8.5.1 Difference Between the Two Query Languages ... 159 8.5.2 Conclusion ... 160 9 User Feedback ... 163 10 Relevant Research ... 169 10.1 Overview Maps ... 169 10.2 Visualizing Hierarchical Structures ... 176 10.3 Familiar Metaphors for Accessing Information ... 179 10.4 Visual Query Languages ... 181 11 Applications ... 187 12 Future Research ... 197 13 Conclusion ... 203 14 Epilogue ... 207 Bibliography ... 211 Appendix: Tutorial ... 217