Downloadable Save-State Which Can Be Loaded Into an Emulator
Total Page:16
File Type:pdf, Size:1020Kb
UC Santa Cruz UC Santa Cruz Electronic Theses and Dissertations Title Searching Within and Across Games Permalink https://escholarship.org/uc/item/3rj5b1jr Author Anderson, Barrett Publication Date 2020 License https://creativecommons.org/licenses/by/4.0/ 4.0 Peer reviewed|Thesis/dissertation eScholarship.org Powered by the California Digital Library University of California UNIVERSITY OF CALIFORNIA SANTA CRUZ SEARCHING WITHIN AND ACROSS GAMES A dissertation submitted in partial satisfaction of the requirements for the degree of DOCTOR OF PHILOSOPHY in COMPUTATIONAL MEDIA by Barrett Anderson December 2020 The Dissertation of Barrett Anderson is approved: ________________________ Professor Noah Wardrip-Fruin, chair ________________________ Professor Adam M. Smith ________________________ Professor Sri Kurniawan ________________________ Henry Lowood, Ph.D. _____________________________ Quentin Williams Acting Vice Provost and Dean of Graduate Studies Copyright © by Barrett Anderson 2020 Table of Contents List of Figures xi List of Tables xvi Abstract xvii Chapter 1. Introduction 1 Contributions 6 Videogame Moments in STEAM Education 6 Videogame Moments and Information Retrieval 10 Research Questions 11 Research Paradigms 12 Chapter 2. Background 14 How Search Engines Work 14 Vector Space Model for Information Retrieval 18 Visualizing High-Dimensional Information 21 Quantitative Evaluation of Information Retrieval Systems 23 Videogame Search 25 The Game Metadata and Citation Project (GAMECIP) 26 Game Moments as Primary Sources 27 iii Chapter 3. Requirements Analysis for Videogame Moment Search 28 Project Summary 28 Research Questions 29 Related Work 31 HCI of IR 32 Indexing Moments in Games 33 Rich-media Search Engines 35 Interview Methods 36 Recruitment 36 Interview Protocol 39 Anchoring Demonstration 40 Analysis 41 Findings 43 Speedrunners 44 Speeedrunner Interview Subjects 44 What kinds of videogame moments do speedrunners seek? 45 How do speedrunners want to identify those moments in a search engine? 46 iv What will speedrunners do with retrieved moments? 47 Streamers and YouTubers 48 Streamer and YouTuber Interview Subjects 49 What kinds of videogame moments do streamers and YouTubers seek? 50 How do streamers and YouTubers want to identify those moments in a search engine? 50 What will streamers and YouTubers do with retrieved moments? 51 Educators 52 Educator Interview Subjects 52 What kinds of videogame moments do educators seek? 53 How do educators want to identify those moments in a search engine? 53 What will educators do with retrieved moments? 54 Developers and Designers 55 Developer and Designer Interview Subjects 56 What kinds of videogame moments do developers and game designers seek? 57 v How do developers and game designers want to identify those moments in a search engine? 57 What will developers and game designers do with retrieved moments? 58 Scholars 59 Scholar Interview Subjects 59 What kinds of videogame moments do scholars seek? 60 How do scholars want to identify those moments in a search engine? 60 What will scholars do with retrieved moments? 61 Findings Summary 63 Technical Recommendations 64 New Platforms 65 Textual Queries and Domains-specific Vocabulary 65 Videos in the Wild 66 Personal Data Sources 67 Research Recommendations 68 Contributions 69 Chapter 4. Exploring Personal Gameplay Videos 71 Project Summary 71 vi Prior Work 72 Generating Videogame Moment Vectors 72 Videogame Moment Space Visualization 74 Search-by-Analogy Proof of Concept 75 Analogy in Cognition and Computation 76 Cognitive Modeling 77 Cognitive Models of Analogy 78 Search-by-Analogy for Videogame Moments 80 Search-by-Analogy Interactive Explanation 82 Gameplay Video Explorer System 87 Possible use patterns 88 Possible gameplay video data sources 89 Other data sources 90 Within-domain analogies 91 Cross-domain analogies 92 Quantitative Evaluation 95 Future Work 101 Contributions 102 vii Chapter 5. Undergraduate Games Corpus 103 Project Summary 103 Prior Work 104 Student Writing Corpora 104 Games Corpora 105 Student Games Corpora 106 Critical Information Theory 107 Data Collection Ethics 109 Corpus Construction 109 Game Collection Process Guidance 110 Adding to the UC eScholarship Archive 111 Corpus Characterization 112 Corpus Derived Projects 117 Future Work 118 Chapter 6. Story Navigator: Retrieval Technology for Interactive Fiction 120 Project Summary 120 Prior Work 121 Interactive Fiction 121 viii Story Content Analysis 122 Story Navigator System 122 Quantitative Evaluation 126 Contributions 131 Future Work 131 Chapter 7. Conclusion 135 Primary Contributions 135 Revisiting Research Questions 136 Future Work 138 Technical Systems 138 Expert/Stakeholder Evaluation 139 Classroom Integration 139 Outlook 140 References 141 Appendix - Undergraduate Game Collection Guide 160 Overview 160 Ethics Note 160 Course Integration 160 ix Game Archiving Forms 161 Linking to Canvas Game Data and preparing for eScholarship Upload 162 Appendix - Study Proposal: Videogame Search in the Classroom 163 Project Overview 163 Classroom Integration 164 Technical Games Research Guest Lectures 164 Focus Group Evaluation 165 Classroom Study Design 166 Proposed Evaluation 167 Expected Findings and Contributions 168 Appendix – Videogame Search Requirements Analysis Interviews 169 x List of Figures Figure 1 Dissertation projects positioned in relationship to human-centered and computational systems research methods, and educational integration. .................... 13 Figure 2 Documents represented as vectors, which can be thought of as points in an n- dimensional space. ................................................................................................... 18 Figure 3. Query vector based retrieval. .................................................................... 19 Figure 4. Semi-structured interview protocol from videogame moments requirements analysis interviews. Q marks questions for subjects, P priming demonstrations intended to guide discussion, and D subject-specific discussion. Items are paraphrased here for brevity. Reproduced from Anderson & Smith, 2019................................................. 39 Figure 5. Hypothetical examples of the visible and invisible features of a videogame moment from Super Mario World (1990). Along with visual information, game systems information such as this is encoded in the videogame moment vector, but not necessarily these specific features. ........................................................................... 74 Figure 6. Videogame Moment Space visualization (Zhang, 2018) for moments from several playthroughs of the game Super Mario World (Tezuka et al., 1991). ............ 75 Figure 7. Example search-by-analogy query and result images for Super Mario World (left) and Super Metroid (right). Moments A, B, and C are the query moments set by the user, and D is the moment returned by the system to complete the analogy (A is to B as C is to D). ........................................................................................................ 81 xi Figure 8. Python code for finding the cosine similarity of the query vector (derived from example times) to all other moments in the corpus. ......................................... 82 Figure 9. An early stage interactive prototype for search-by-analogy in videogame moments built in Python with Jupyter Widgets. Elements include screenshots for moments A, B, C, and D (top), sliders allowing a user to set moments (middle), and a graph displaying cosine similarity for all moments (bottom). ................................... 85 Figure 10. Step-by-step illustrations search-by-analogy for videogame moments from my explainer, preceding the interactive element. The first graph (1) depicts moment vectors for hypothetical videogame moments A and B. The next (2) depicts the B-A vector of the difference between those moments. The third graph (3) illustrates the addition of this B - A difference vector to C to form a synthetic moment vector, which then (4) becomes the query vector. The fifth graph (5) illustrates how a search system finds the best match for the query vector from the set of moments that exist. The final graph (6) exchanges the arbitrary axes for ones that provide meaningful information (similarity to moment C and to the difference vector), in a way that matches the interactive visualization. .......................................................................................... 86 Figure 11. Search-by-analogy explorable explanation interactive graph. The user selected points A, B, and C are highlighted, and the point chosen by the system as the best search-by-analogy match (the point with the highest cosine similarity to the (B- A)+C vector ) is highlighted as point D. Mouseover of the points on the graph overlays the highlighted moment’s screenshot and similarity scores. ..................................... 87 xii Figure 12. An early version of the Gameplay Video Explorer site with moments from a speedrun of Super Mario World (Tezuka et al., 1991) loaded, displaying a search-by- analogy style graph (B. R. Anderson et al., 2019). ................................................... 93 Figure 13. The Gameplay Video Explorer site, with data from two videos loaded (a speedrun of Super Mario World, and of Super Mario Bros.). Moments from each video are mapped to the same space, and color-coded by source. Mouseover reveals the specific moment. In order to handle diverse video lengths and still remain performant, a sampling approach is