Visualization 논문 리뷰 작업 중에 논문을 유형별로 나눌 필요성이 느껴져서 찾다가, 블로그에 정리해 두면 좋을거 같아서 IEEE VIS 2015 학회에 정의된 논문 제출 가이드라인 일부를 그대로 영문 번역하였다. 발번역임. (VAST 2016 에서 약간 수정된 논문 타입을 정의하였기에 추가 번역하였음.)
http://ieeevis.org/year/2015/info/call-participation/paper-submission-guidelines
http://ieeevis.org/year/2016/info/call-participation/vast-paper-types
VAST 2016 Paper Types
VAST에는 TVCG 트랙과 Conference-only 트랙, 두개의 트랙이 있는데, 각 트랙은 서로 다른 레벨의 originality, rigor, 그리고 significance를 가진다. 보통 VAST 논문들은 다른 두 VIS 학회 - InfoVIS, SciVIs - 의 논문들처럼 상세한 접수 가이드라인에 따라 작성, 접수, 리뷰의 과정을 거친다. 그러나, visual analytics의 빠른 과학, 기술, 응용의 발전에 따라, 수시로 VAST 출판물에 대한 이해도를 조정하는 것이 현명하다. 기존 5개 형식의 논문 타입에 대한 논의는 뒤로 하고, 아래에 VAST 2016의 논문 타입에 대해 명시하였다.
Visual analytics, concepts, theories, algorithms, techniques, designs, systems, empirical studies 와 applications 는 일반적으로 분석, 시각화, 상호작용이 인간과 기계간 조화를 최적화 하도록 통합되는 곳에 컨텍스트를 생성한다. 이 컨텍스트는 VIS의 다른 학회들과 차별된다: 데이터가 spatial 인지 non-spatial인지, 기술들이 human-centric 인지 machine-centric 인지, application domain 이 대부분 any academic discipline, industry, business sector, 혹은 governmental operation 인지. 이러한 맥락에서, 개개의 VAST 논문은 새로운 기여가 존재하는 곳에 강하게 포커스를 맞추거나 다른 측면의 통합을 강조한다.
VAST has two tracks, TVCG and Conference-only tracks, which correspond to different levels of originality, rigor, and significance. In general, VAST papers should be written, submitted and reviewed in the same way as papers at the other two VIS conferences (i.e., InfoVis and SciVis), following the detailed submission guidelines. However, with the rapid development of the science, technology and application of visual analytics, it is sensible to adjust our understanding of VAST publications from time to time. We provide the following clarifications about paper types for VAST 2016, beyond the discussion of the five paper types in the shared guidelines.
In visual analytics, concepts, theories, algorithms, techniques, designs, systems, empirical studies and applications normally create a context where analysis, visualization and interaction are integrated to optimize the combination of human and machine capabilities. It is this context that differentiates VAST from other conferences in VIS, while data involved can be spatial or non-spatial, techniques can be human-centric or machine-centric, and the application domain can be almost any academic discipline, industry, business sector, or governmental operation. Within such a context, an individual VAST paper may give a strong focus on an aspect where novel contributions reside, or place its emphasis on the integration of different aspects.
VAST 논문은 일반적으로 6개의 카테고리로 분류된다:
이론과 모델
기술과 알고리즘
디자인 연구
실증 연구 (기존 가이드에서 Evaluation 형태로 분류되는 논문들.)
시스템
응용 (Design Study 에서 별도로 분류된 카테고리.)
VAST papers typically fall into one of these six categories:
Theory and Model
Technique and Algorithm
Design Study
Empirical Study (referred to as the Evaluation type in in the shared guidelines)
System
Application (as a separate category from the Design Study type in the shared guidelines)
Theory and Model
Fundamentals of visual analytics. Conceptual understanding and modelling of visual analytics (e.g., definitions, taxonomies, analytic frameworks, and research methods, etc.). Philosophical and sociological discourses of visual analytics (e.g., human vs machine, ethics, data security, uncertainty, biases, and privacy, etc.). Perception and cognition in visual analytics.
Mathematical abstraction and modelling of visual analytics processes. Concepts and models that govern quality metrics and benchmarks for evaluating visual analytics processes and systems.
Examples:
Megan Monroe, et al. "Temporal event sequence simplification." VAST 2013 Honorable Mention.
Technique and Algorithm
Visualization techniques in visual analytics processes. Close integration of technical components of visual analytics (e.g., statistical analysis, human-defined and machine-learned algorithms, knowledge representations, visualization/interaction techniques and methodologies, etc.) for supporting visual data mining.
Visual analytics for supporting the advancement of non-visual technical components of visual analytics (e.g., visual analytics for supporting model development, simulation, learning, monitoring, and optimization).
Integrated data acquisition, management, retrieval, processing and transformation in visual analytics (e.g., multi-sources; multi-resolution; data provenance; uncertainty; real world measures; textual, audio, visual and other media; factual, statistical, semantic, synthesized, and hypothesized data; etc.).
VA techniques for spatial and non-spatial data, temporal data, streaming data, quantitative and qualitative data, text and document data, and so on. Techniques for production, presentation, and dissemination of VA results.
Examples:
Thomas Muhlbacher and Harald Piringer. “A Partition-based framework for building and validating regression models.” VAST 2013 Best Paper.
Stef van den Elzen et al. “Reducing snapshots to points: a visual analytics approach to dynamic network exploration.” VAST 2015 Best Paper.
Empirical Study
Understanding human-centric components in visual analytics processes (e.g., perception, cognition, interaction, communication, collaboration, etc.).
Understanding human capabilities and limitations in data intelligence (e.g., exploration, navigation, sensemaking, context awareness, knowledge discovery, learning, argumentation, causality reasoning, accountability, biases, etc.).
Understanding visual signatures in data intelligence (e.g., patterns of clusters, patterns of anomalies, etc.).
Understanding the potential merits and demerits of technologies in visual analytics (e.g., display technologies, interactive technologies, automated analytics, crowdsourcing analytics, and so on).
Human-centric comparative studies on aspects of visual analytics (e.g., visual representations, interaction techniques, active learning, visual analytics literacy, requirements analysis, etc.).
Evaluation methodologies for visual analytics techniques and systems in real world environments.
Different quantitative and qualitative (including ethnographic) forms of empirical studies (e.g., lab-based studies, field studies, crowdsourcing, group discussions, surveys, interviews, user experience observation, shadowing, case studies and casebook construction, etc.)
Transformation of scenarios and data captured in studies to benchmark problems and data-driven metrics.
Examples:
Narges Mahyar and Melanie Tory. "Supporting communication and coordination in collaborative sensemaking." VAST 2014 Best Paper.
Hua Guo et al. "A case study using visualization interaction logs and insight metrics to understand how analysts arrive at insights." VAST 2015 Honorable Mention.
Design Study
Designing disseminative visual analytics (e.g., storytelling, illustration and animation, public engagement, etc.)
Designing observational visual analytics (e.g., multivariate data, streaming data, multimedia data, geospatial data, spatio-temporal, etc.)
Designing analytical visual analytics (e.g., clustering, anomaly detection, association and network analysis, correlation, causality, uncertainty, etc.)
Designing model-developmental visual analytics (e.g., exploring parameter space, and model space, supporting dimensionality reduction and machine learning, model-developmental life cycle, etc.)
Design methodologies for real world visual analytics systems and users.
Examples:
Jian Zhao et al. "#FluxFlow: Visual analysis of anomalous information spreading on social media." VAST 2014 Honorable Mention.
System
Methodologies for engineering real world visual analytics systems.
System platforms (from wearable devices to desktops to large infrastructures, and from architectures and software libraries (toolkits), to stand alone systems and apps, to online services and open source repositories).
Comparative studies on real world visual analytics systems.
Development tools for the software lifecycle of visual analytics systems, including requirements analysis, system specification, system design, system implementation, system testing, user evaluation, and system maintenance).
Addressing challenges in real world visual analytics systems (e.g., provenance management, scalability, uncertainty, open testbeds, etc.).
Automation, customization, and personalization, and interoperability.
Best practices (e.g., interoperability, workflow design, cost-benefit analysis, standardization, etc.)
Examples:
Tanja Blascheck et al. "VA2: A visual analytics approach for evaluating visual analytics applications." VAST 2015 Honorable Mention.
Application
Delivering visual analytics solutions to applications in academic disciplines (e.g., physical sciences, biological and medical sciences, engineering sciences, social sciences, arts and humanities, and sports sciences).
Delivering visual analytics solutions to applications in industries and governance.
Delivering visual analytics solutions to applications in public services and entertainment (e.g., resilience, healthcare, transport, sports, tourism, broadcasting, and social media).
Examples:
Conglei Shi, et al. "LoyalTracker: Visualizing loyalty dynamics in search engines." VAST 2014 Honorable Mention.
VIS 2015 Paper Types
VIS 논문의 유형은 대게 5개 카테고리로 나누어진다: technique, system, design study, evaluation, or model. 이 카테고리들에 대해 아래에 간단히 설명하였다. 논문 유형은 논문 제출 과정 중에 정해지긴 하지만, 한개 이상의 카테고리를 가질 수 있다. "Process and Pitfalls in Writing Information Visualization Research Papers" by Tamara Munzner 를 참고하여 어떻게 VIS 논문을 잘 쓸 수 있는지 참고하라.
A VIS paper typically falls into one of five categories: technique, system, design study, evaluation, or model. We briefly discuss these categories below. Although your main paper type has to be specified during the paper submission process, papers can include elements of more than one of these categories. Please see "Process and Pitfalls in Writing Information Visualization Research Papers" by Tamara Munzner for more detailed discussion on how to write a successful VIS paper.
Technique 눈문
이 유형의 논문은 해당 학계에서 이전에 본적 없던 새로운 기술(techniques)이나 알고리즘을 소개하거나, 기존 기술과 알고리즘을 현저하게 발전시킨 경우로, 예를 들어 이전보다 훨씬 큰 데이타셋으로 확장하거나 더 많은 방면으로 사용하게 되는 기술로 일반화 하는 경우이다. 기술이나 알고리즘의 상세 설명이 완전히 제공되어서 실력을 갖춘 대학원생이 이를 구현할 수 있어야 하고, 저작자는 프로토 타입 구현물을 만들어야 한다. 관련 이전 논문 참조가 반드시 있어야 하고, 이전 방법을 뛰어넘는 새로운 방법의 장점이 명백하게 기술되어야 한다. 이 새 방법이 적절한지, 제한이 있는지에 대해 작업과 데이타셋의 논의가 필요한다. 비공식 또는 공식적인 user studies 나 다른 방법을 통한 평가(Evaluation)는 논문을 더 탄탄하게 만들어 주지만, 필수사항은 아니다.
Technique papers introduce novel techniques or algorithms that have not previously appeared in the literature, or that significantly extend known techniques or algorithms, for example by scaling to datasets of much larger size than before or by generalizing a technique to a larger class of uses. The technique or algorithm description provided in the paper should be complete enough that a competent graduate student in visualization could implement the work, and the authors should create a prototype implementation of the methods. Relevant previous work must be referenced, and the advantage of the new methods over it should be clearly demonstrated. There should be a discussion of the tasks and datasets for which this new method is appropriate, and its limitations. Evaluation through informal or formal user studies, or other methods, will often serve to strengthen the paper, but are not mandatory.
System 논문
이 유형의 논문은 알고리즘, 기술적 요구사항, 사용자 요구사항, 설계 등을 융합하여 주요 문제를 풀어낸다. 제시된 System 은 새롭고 중요한 것으로서, 구현 된 것으로 여겨진다. 중요한 설계 결정을 위한 이론이 뒷받침 되고, 문서화 되어 이미 사용중인 동종 최고의 시스템과 비교되어진다. 이 비교에는 몇 가지 중요한 측면에서 이 시스템이 다른 시스템들과 어떻게 다른지, 그리고 무엇이 우수한지에 대한 구체적인 논의가 포함된다. 예를 들어, 기술된 시스템은 시각화(visualization) 시스템의 성능이나 사용성에서 상당한 우수성이나 새로운 가능성을 제공할 수도 있다. 외부적인 요소들은 (프로세서 성능, 메모리 크기, 혹은 OS의 기능면에서의 향상 등) 이러한 비교에 영향을 미치므로 가능한 모두 제거해야 한다. 추가 제안으로는, "How (and How Not) to Write a Good Systems Paper" by Roy Levin and David Redell 과 "Empirical Methods in CS and AI" by Toby Walsh 를 훑어볼 것을 권장한다.
System papers present a blend of algorithms, technical requirements, user requirements, and design that solves a major problem. The system that is described is both novel and important, and has been implemented. The rationale for significant design decisions is provided, and the system is compared to documented, best-of-breed systems already in use. The comparison includes specific discussion of how the described system differs from and is, in some significant respects, superior to those systems. For example, the described system may offer substantial advancements in the performance or usability of visualization systems, or novel capabilities. Every effort should be made to eliminate external factors (such as advances in processor performance, memory sizes or operating system features) that would affect this comparison. For further suggestions, please review "How (and How Not) to Write a Good Systems Paper" by Roy Levin and David Redell, and "Empirical Methods in CS and AI" by Toby Walsh.
Application / Design Study 논문
이 유형의 논문은 한 응용분야에서 시각화나 시각 분석 기술을 적용할 때 발생되는 선택 사항들을 탐구한다. 예로, 대상 업무의 요구사항에 대한 상호작용 기술들과 시각적 인코딩 등을 들 수 있다. 유사하게, Application 논문은 공학 과학 문제들에 대한 통찰들을 모으기 위해 사용되는 시각화 기술들을 설명할 때, 표준으로 사용되어져 왔다. 그럼에도 불구하고 상당한 양의 도메인 배경 지식 application이 목적 업무의 상세사항을 논하는 콘텍스트의 프레이밍을 제공함에도 불구하고, case study의 주요 초점은 시각화 내용이어야만 한다. 응용 도메인에서 생성된 통찰들을 포함한 Application/Design Study 논문의 결과물들은 분명하게 전달되어야 한다. 당면한 문제를 풀기위한 새로운 기술과 알고리즘을 기술하는 것은 design study 논문을 강화하지만, Technique 논문만큼 참신함이 크게 요구되어지진 않는다. 필요할 시, 기본적인 파라메트릭 스페이스와 그것의 효과적인 검색에 대해 적절한 설명으로 밝혀야 한다.
Application / Design Study papers explore the choices made when applying visualization and visual analytics techniques in an application area, for example relating the visual encodings and interaction techniques to the requirements of the target task. Similarly, Application papers have been the norm when researchers describe the use of visualization techniques to glean insights from problems in engineering and science. Although a significant amount of application domain background information can be useful to provide a framing context in which to discuss the specifics of the target task, the primary focus of the case study must be the visualization content. The results of the Application / Design Study, including insights generated in the application domain, should be clearly conveyed. Describing new techniques and algorithms developed to solve the target problem will strengthen a design study paper, but the requirements for novelty are less stringent than in a Technique paper. Where necessary, the identification of the underlying parametric space and its efficient search must be aptly described. The work will be judged by the design lessons learned or insights gleaned, on which future contributors can build. We invite submissions on any application area.
Evaluation 논문
이 유형의 논문은 사람에 의한 시각화와 시각적 분석의 사용을 탐구하고, 일반적으로 시각화 기술이나 시스템의 경험기반 연구가 제공된다. 저작자는 이러한 스터디가 자체적으로 사용되어지는 시스템을 구현할 필요는 없다; 연구의 기여도는 연구중인 시스템 또는 기술의 참신함과는 반대로 유효성과 경험적으로 얻은 결과의 중요성에 의해 평가된다. 학회의 커미티는 엄격한 실험을 설계하고 수행하는 중요성과 어려움을 높이 평가한며, 적절한 가설(hypotheses), 업무, 데이타셋, 대상의 선택, 측정, 유효성과 결론 등에 대한 것들이 포함된다. 이러한 노력은 실험에 대한 단순 기술보다는 예측과 설명을 목표로 한다. 인간을 대상으로 하는 실험의 설계에 익숙하지 않은 저작자는, 엄격한 경험적 프로토콜의 설계와 결과자료의 통계적 분석에 경험이 있는 심리학이나 human-computer 상호작용 영역의 동료를 원할 수 있다. 다른 새로운 형태의 evaluation 또한 권장된다.
Evaluation papers explore the usage of visualization and visual analytics by human users, and typically present an empirical study of visualization techniques or systems. Authors are not necessarily expected to implement the systems used in these studies themselves; the research contribution will be judged on the validity and importance of the experimental results as opposed to the novelty of the systems or techniques under study. The conference committee appreciates the difficulty and importance of designing and performing rigorous experiments, including the definition of appropriate hypotheses, tasks, data sets, selection of subjects, measurement, validation and conclusions. The goal of such efforts should be to move from mere description of experiments, toward prediction and explanation. We do suggest that potential authors who have not had formal training in the design of experiments involving human subjects may wish to partner with a colleague from an area such as psychology or human-computer interaction who has experience with designing rigorous experimental protocols and statistical analysis of the resulting data. Other novel forms of evaluation are also encouraged.
Theory/Model 논문
이 유형의 논문은 시각화와 시각적 분석에 대한 기초적 이론의 새로운 해석을 제시한다. 구현은 보통 이러한 논문과는 연관이 없다. 논문은 어떻게 시각화 기술이 보완되고 인간의 시각과 인지의 요소들을 활용하는지에 대한 우리의 이해 속에서 기본적인 발전성에 초점을 맞춰야 한다.
Theory/Model papers present new interpretations of the foundational theory of visualization and visual analytics. Implementations are usually not relevant for papers in this category. Papers should focus on basic advancement in our understanding of how visualization techniques complement and exploit properties of human vision and cognition.
Prev
