본 글은 아래의 논문에 대한 개인적인 요약 리뷰입니다. 의역과 주관적인 해석이 담긴 리뷰이므로 참고 하여주시기 바랍니다.

Lee, B., Plaisant, C., Parr, C. S., Fekete, J. D., & Henry, N. (2006, May). Task taxonomy for graph visualization. In Proceedings of the 2006 AVI workshop on BEyond time and errors: novel evaluation methods for information visualization(pp. 1-5). ACM.



논문의 제목은 Task taxonomy for graph visualization. 이를 제목과 같이 '그래프 시각화 자료를 읽을 때의 작업 분류' 라고 의역하였으나 의미 전달이 모호해서 본문에서는 'task'를 '작업'으로 번역하지 않도록 하겠다.

참고로 여기서 말하는 Graph visualization이란 노드와 링크가 있는 그래프를 말한다.


본 논문을 읽기 전에 참고해야 할 논문이 있는데, R. Amar외 2명이 저술한 Low-Level Components of Analytic Activity in Information Visualization를 미리 읽어두면 도움이 된다. 그 이유는 본 논문에서 분류한 task를 R. Amar가 제시한 LOW-LEVEL TASKS로 나누어 설명하기 때문이다. R. Amar가 제시한 LOW-LEVEL TASKS는 정보 시각화에 대해 광범위하게 사용될 수 있는 분류 방식이고, 본 논문은 이를 Graph visualization라는 특정한 주제에 적용시킨 논문이라고 할 수 있다.


아래의 표는 R. Amar가 제시한 LOW-LEVEL TASKS에 본 논문에서 제시한 세가지 task를 추가한 것이다.(Scan, Set Operation, Find Adjacent Nodes)

(이후 첨부한 표들의 출처는 모두 : http://www.infovis-wiki.net/index.php?title=Tasks_Taxonomy_for_Graphs)

LOW-LEVEL TASKS

TaskDescription
Retrieve ValueGiven a set of cases, find attributes of those cases.
FilterGiven some conditions on attributes values, find data cases satisfying those conditions.
Compute Derived ValueGiven a set of data cases, compute an aggregate numeric representation of those data cases.(e.g. average, median, and count)
Find ExtremumFind data cases possessing an extreme value of an attribute over its range within the data set.
SortGiven a set of data cases, rank them according to some ordinal metric.

Determine Range Given a set of data cases and an attribute of interest, find the span of values within the set.

Characterize DistributionGiven a set of data cases and a quantitative attribute of interest, characterize the distribution of that attribute’s values over the set.
Find AnomaliesIdentify any anomalies within a given set of data cases with respect to a given relationship or expectation, e.g. statistical outliers.
ClusterGiven a set of data cases, find clusters of similar attribute values.
CorrelateGiven a set of data cases and two attributes, determine useful relationships between the values of those attributes.
ScanQuickly review a set of items.
Set OperationGiven multiple sets of items, perform set operations on them. For example, find the intersection of the set of nodes.

Find Adjacent Nodes

Given a node, find its adjacent nodes.


위의 LOW-LEVEL TASKS를 바탕으로, 본 논문에서는 Graph visualization을 읽을 때의 task를 크게 네 가지로, 그리고 분류에 포함되지 않는 예외적인 task들을 따로 명시했다.


Examples 열에 있는 괄호 안의 단어는 다음을 뜻한다.

Examples are illustrated using 4 types of graphs:

  • (FOAF): friend-of-a-friend
  • (FW): food web
  • (GO): gene ontology
  • (ARM): airport routing map


Topology-based Tasks

TaskDescriptionExamples
Adjacency (direct connection)
  • Find the set of nodes adjacent to a node?
  • How many nodes are adjacent to a node?
  • Which node has a maximum number of adjacent nodes?
  • (FOAF) Find the names of the direct friends of Eric.
  • (FW) How many kinds of organisms do golden eagles eat?
  • (FOAF) Who is the most popular person?
Accessibility (direct or indirect connection)
  • Find the set of nodes accessible from a node.
  • How many nodes are accessible from a node?
  • Find the set of nodes accessible from a node where the distance is less than or equal to n.
  • How many nodes are accessible from a node where the distance is less than or equal to n?
  • (FOAF) Who are your friends, your friends’ friends, and so on?
  • (FOAF) How many friends are you connected to in this way?
  • (ARM) To what cities can we go from Seoul, Korea by changing planes only once?
Commmon Connection
  • Given nodes, find a set of nodes that are connected to all of them.
  • (FOAF)Find all the people who know both John and Jack.
Connectivty
  • Find the shortest path between two nodes.
  • Identify clusters.
  • Identify connected components.
  • Find bridges.
  • Find articulation points.
  • (ARM) What is the shortest path from Seoul, Korea to Athens, Greece?
  • (FOAF) Count the number of clusters.
  • (FW) There may be subgraphs independent of each other. Count the number of connected components in the graph.
  • (FOAF) Who is the person whose removal from the graph results in an unconnected graph?
  • (FW) Which is the eating link whose removal from the graph results in an unconnected graph?



Attribute-based Tasks

TaskDescriptionExamples
On the Nodes
  • Find the nodes having a specific attribute value.
  • Review the set of nodes.
  • (FOAF) Who do you know from the people currently shown on screen?
  • (FOAF) How many people do you know from the ones currently shown on screen?
  • (FOAF) Are there any foreign-sounding names?
On the Links
  • Given a node, find the nodes connected only by certain types of links.
  • Which node is connected by a link having the largest/smallest value?
  • (GO) Find the nodes connected by “is-a” relationships from the “Biological Process” node.
  • (FW) If a link has an attribute representing the percentage of the diet, what is main food of American crow?

 


Browsing Tasks

TaskDescriptionExamples
Follow Path
  • Follow a given path
  • (FOAF) A user looks into A’s friend B, B’s friend C, and C’s friend D.
  • (FW) Follow the flow of energy from grasses, to a rabbit that eats grass, to a carnivore that eats the rabbit, and to a carnivore that eats that carnivore.
Revisit
  • Return to a previously visited node.
  • (FOAF) After they follow a path in the above task, they may want to see A’s other friends.
  • (FW) Find another carnivore that eats the same rabbit.


Overview Tasks

This is a compound exploratory task to get estimated values quickly. For example, we might ask users to estimate the size of the social network. Note that sometimes it is more important to be able to estimate the answer than to get an accurate one. Some of the topology tasks can be done easily using an overview of the graph as well. For example, using particular layout algorithms, it is easy to see whether or not there are clusters and connected components. Other algorithms help to find shortest paths between nodes. Overview also helps to find patterns.

Examples:

  • estimate size of the network
  • estimate the number of connected components
  • is the network clustered?
  • can you identify different patterns of connection?
  • (FOAF) has the network a small-world structure?


High-Level Tasks

High-Level tasks which are not a combination of lower level tasks.

  • When we compare two or more food webs, we can ask the following questions: What do they have in common? What are the differences among those food webs? Is there any missing or conflicting information?
  • Due to errors in the data, several nodes may represent the same entity. For example, the co-authorship graphs often have duplicate author nodes. One important task is to identify whether two or more nodes represent the same person.
  • How has the graph changed over time?



이 논문에서는 단순히 분류 기준만을 제시했지만, 좀 더 나아가 각 task에 대해 사용자 평가를 통한 수행 시간, 피로도 등을 측정 해 보고,

이를 통해 효과적인 Graph visualization 방법에 대한 가이드라인을 제시할 수 있지 않을까 생각해본다.


'논문 리뷰' 카테고리의 다른 글

어떻게 시각화(Visualization) 논문을 읽을까  (0) 2016.09.11

본 글은 아래의 논문에 대한 개인적인 요약 리뷰입니다. 의역과 주관적인 해석이 담긴 리뷰이므로 참고 하여주시기 바랍니다.

Laramee, R. S. (2011). How to read a visualization research paper: Extracting the essentials. IEEE computer graphics and applications31(3), 78-82.



이 논문은 시각화(Visualization) 연구주제를 다룬 논문을 어떻게 효과적으로 읽을 수 있을 지에 대한 방법을 제시.


시각화 관련 논문을 읽을 때의 네 가지 중요 정보 :

1. 컨셉(The Concept) : 연구 목적이 무엇인가, 새로운 것은 무엇인가

2. 구현(The Implementation) : 어떻게 구현했는가, 어떻게 가설 검증하였는가

3. 관련 연구(Related Work) : 이전에 선행되었던 관련된 연구는 어떤 것인가

4. 자료 특성(Data Characteristics) : (수행한 실험에 대한 결과 값의) 차원(2D, 3D, ..), 

시간에 따라 변하는 또는 변하지 않는 데이터 값인지(time-dependant or static), 

해상도(측정한 도구의 최소 눈금 단위가 어떠한지), 

벡터량인지 스칼라 양인지 등등


컨셉(Concept)과 구현(Implementation)의 차이점 :

컨셉은 추상적인 아이디어에 그치는데 반해, 구현은 대체로 물리적으로 측정 가능한 실체가 있다.

(보통 구현 이후에는 정량적인 실험으로 이어지는게 일반적)

또한 하나의 컨셉에 대한 여러 다양한 구현 방법이 있을 수 있다.



예시)


제목 : Fast and Resolution Independent Line Integral Convolution by D. Stalling and H.-C. Hege


1. 컨셉 : 이 논문은 the Line Integral Convolution (LIC) algorithm의 더 빠른 버전을 제시

2. 구현 : streamline computations의 중복을 줄이는 방법,  streamline integrator with adaptive stepsize control 이용

3. 관련 연구 : This work builds upon and improves the original LIC algorithm

4. 자료 특성 : spatial dimensionality: 2D

temporal dimensionality: steady (static)

resolution: uniform

grid structure: uniform resolution, regular

data type: vector



'논문 리뷰' 카테고리의 다른 글

그래프 시각화 자료를 읽을 때의 작업 분류  (0) 2016.09.18

+ Recent posts