Expressiveness And Effectiveness Of The Visualization Computer Science Essay
Visualization is a method or a transformation of data or information into images, diagrams, or animations. Concise Oxford Dictionary states that visualization means “to imagine or remember as if actually seeing”. Besides, in Webster’s Ninth New Collegiate Dictionary, it has defined visualization as “the act or process of interpreting in visual terms or of putting into visual forms”. [1] In another words, visualization is the communication of information using graphical representations.
There is no longer an obstacle for collecting data or information though extracting necessary values from collected information has turned out to be gradually more complex and complicated. Since early days before the written language has formalized, we have been using pictures for communication and visual imagery has been an efficient way to correspond both tangible and abstract ideas. We, humans, have complex and great vision system which we utilize and rely for everything we do on a daily basis while the speed of analyzing text is quite limited for us by the sequential process of reading.
Purpose of Visualization
The main reason of visualization is to convey, explore and analyze information. To be more specific, visualization is used to present large amount of information compactly from various view points and at several levels of details. Furthermore, it helps us extract the important information which is hidden within the data. Visualization is essential to manage today’s world information of computers, satellites, digitized systems and etc.
Some data sets are naturally better to be represented visually since we possess the abilities of powerful human vision system. Graphical representation makes easier for analyzing the data especially when all the information and its relation are segregated with different colors, shapes, and size. [1] As today’s technology is tremendously growing and with the inventions of all the computers and their capabilities to generate large data sets, visualization is the most suited technology to extract and study the information from collected raw data.
Some examples of visualization comprise mapping the blogosphere, web trend map which is a detailed study of the current online trends, and hierarchical structure of the internet that displays the connectivity and how it is being managed. In addition, visualization offers considerable financial advantages in today’s competitive world. Computer simulation together with visualization can save product expenses and time required for production.
Types of Visualization
There’re few terminologies which can be used to represent visualization. Scientific visualization in computer science field means the method of graphically displaying real or simulated data. It is a fundamental process in the innovative realization of scientific ideas and its basic visualization techniques contain surface rendering, volume rendering, animation, processing algorithms and other sensory presentation such as sound or touch. [1]
Another phrase to express visualization is data visualization. Data visualization is more general compared to scientific visualization as its data sources involve business, marketing and financial data which are beyond sciences and engineering fields. [1] Moreover, it consists of statistical processes and other standard data analysis techniques.
Information visualization is used to visualize abstract information and abstract structures, directory files on computers, hypertext documents on World Wide Web, etc. [1] It draws on the intellectual history of several customs like computer graphics, human computer interaction and statistical graphics.
Visualization can be classified based on context in which data exists. Based on the data sets, the techniques of visualization are differed. Scientific visualization methods are used when data exists up to three spatial coordinates and time dimension whereas information visualization is for data in higher dimensional or abstract spaces. Scientific visualization and information visualization overlap each other and they are allied fields. [1] The relationship of three different visualizations can be found as in figure 1.
Figure 1: Types of Visualization
Visualization Process
Figure 2 illustrates the steps of visualization process. [2] The very first step of designing visualization is about analyzing the data to be visualized. It is necessary to find out whether data from a database, a file or some source, simple or complex, is able to be structured and allows for easy modification to suit its visualization. The designer needs to take note of the presentation of the visualization results and the information the users wish to extract from the enormous data set. Raw data will be then transformed into symbolic representations.
Secondly, the data values themselves or the data attributes are mapped as graphical objects, such as shapes, lines, color, position and size. The last component of visualization process is rendering of graphic objects by the computer onto the display and generation of visualization for the user’s interpretation.
Figure 2: Visualization Process from High Level View
Visualizing Information
One of the fundamental questions in information visualization is how to describe expressiveness and effectiveness, the two mathematical measures of visualization, which can be applied at all stages of visualization process. Besides when visualizing, there’re some important parameters to consider such as visualization and symbols, graphic features and the eight visual variables.
Expressiveness and Effectiveness of the Visualization
Jock D. Mackinlay, an American visualization expert, states “a visualization is expressive if a visualization encodes all relevant information and only that information”. [3] That denotes the person may see all information he/ she wants to examine without any distractions. Therefore, expressiveness measures the concentration of information. Perfect visualization means with ideal expressiveness and it is tough to achieve in reality. Expressing too much information will lead to interference of interpreting essential information and expressing less information will miss out important datasets need to be visualized. Effectiveness means that all information is presented clearly and quickly in a cost effective manner. [3] Hence effectiveness measures a precise cost of information awareness.
Beshers & Feiner, the scientists, adapt these two measures and express it as potential expressiveness and potential effectiveness. [3] A visualization is potentially expressive, if it has the potential and under the user control to display all essential information over time. It is potentially effective, if the information presented is sufficiently clear over time.
Visualization and Symbols
In visualization, symbols create a wide range of new possibilities for visual effects. Symbols have been used to connect with many intentions and they play as valuable roles in information visualization. Visual objects are graphical symbols which are parts of visualization like arrows, labels, dots and etc. To discover relations or patterns of visualization, Cleveland states that there are two major steps. [4] The first step is a mapping between graphic symbols and the represented data. Lastly finding patterns on the screen that imply the patterns in the data.
Graphic Features
Graphics are represented in three or more dimensions. Every single point of a graphic is construed as a relation between two positions x and y with a third variable value z. Graphics can be analyzed in three main steps. [2] First is to perceive groups of objects pre-attentively, followed by characterizing those groups cognitively. The final step is to examine special cases which are out of the group.
The Eight Visual Variables
To represent different aspects of the same information, choosing visual variables is crucial and can affect the perception and understanding of the presented information. Thus, it is essential to understand graphic primitives and their variables. The eight visual variables are as below. [2]
Position
It is the most important visual variable and changes in x, y location. In visualization, the spatial arrangement is the very first thing to be done. That is the reason why positioning has the greatest impact on the display of information.
Shape
Shapes or marks refer to points, lines, areas, volumes, and their compositions, and they are graphic primitives that represent data. There are infinite number of shapes and they are used for categorization.
Size
Size changes in area, length or volume. It influences the way of individual data representation and display.
Brightness
Brightness or luminance is good for large interval and continuous data. However, there is a limitation to distinguish among all those different levels of brightness.
Color
It changes according to two parameters, hue and saturation. It requires mapping of data values to individual color codes.
Orientation
Orientation changes in alignment. It cannot be used for all marks. For example, circles look the same even their orientations change. Shapes with natural single axis are the best to apply orientation.
Texture
It is a combination of many other visual variables including marks, color, orientation and so on.
Motion
Motion describes all visual variables change over time and it can convey more information.
Human Perception System
Visual perception means the ability to interpret and process information from visible light in the surrounding environment. Not everyone perceives data exactly the same. Different viewers differently interpret the identical visual representations.
When designing visualization, to reduce the confusion later on, the designers need to take account of color usage of graphical entities for accurate measurement, quantity of distinct entities, and etc. Besides, we also need to consider the primitives that humans usually detect pre-attentively and the level of accuracy we perceive various primitives. Consequently, when we visualize data, it is a basic requirement to learn the limits of human perception since we need to factor in these limitations and avoid producing images with vague or deceptive information.
Visual System
The human eye is composed of many parts. [5] They obtain visual images, focus them accurately and send messages to the brain. The main sensory component of vision assembles light scattered from objects and forms a two dimensional function on the photoreceptors, the small sensory devices which respond in the presence of photons making up light waves.
Information related to the external objects in the environment is captured through the visual system. Light rays from an object enters through the outer part of the eye, named the cornea. It helps the eye to focus to make things look sharp and clear. Then, the light rays travel towards an opening called the pupil, the dark round circle in the middle of the colored part of the eye. The colored eye is called the iris and the pupil is just a hole in the iris. The iris controls the amount of light goes into the eye. Besides, your eye also possesses a lens to focus the light rays. Light passes through the lens till the back of the eye, the retina. It has millions of tiny light sensitive cells sending messages to parts of the brain, the optic nerves.
Field of View
A pair of normal healthy human eyes can view about 200 degrees horizontally where approximately 120 degrees of which are shared by both eyes and giving rise to what’s known as binocular vision. [6] It has a field of view of 135 degrees vertically. However, as we get older, these values decrease. Both of human eyes are positioned more or less on the front of our heads and it is common in prey species as it helps increase an animal’s total field of view.
Angular Resolution
Angular resolution refers to the minimum distance at which our eyes can differentiate things of the same size and shape from each other. [6] The typical set of human eyes has an angular resolution of one minute of arc. It means objects one degree apart from each other can be distinguished. Therefore, angular resolution is useful when we need to differentiate similar objects. Nevertheless, every human eye is different and their angular resolution varies based on eyesight strength, eye shape and age.
The Blind Spot
The photoreceptor cells in our eyes are used to perceive light and information being received is relayed to the brain via the optic nerve. Blind spot is the visual field where it lacks the light detecting photoreceptor cells on the optic disc of the retina. [6] A small part of the field of vision is not perceived as there are no cells to detect light. Normally, with two eyes, the brain interpolates the blind spot based on surrounding details and information from the other eyes so that the blind spot would not be detected. However, blind spot can be perceived easily with one eye closed.
Perceptual Processing
Attention acts as a critical role in perceiving information. Perception can be pre-attentive or attentive. Usually the flow of perceived information starts from the low level pre-attentive towards the high cognitive stages.
Professor Treisman states perceptions that can be performed in less than 200 to 250ms are regarded as pre-attentive. [7] Initiating random locations of the elements in display by human eyes normally take at least 200ms. That determines attention cannot be pre-focused on any particular situation and information is processed in parallel by the human visual system. Pre-attentive perception requires its objects to possess a unique feature, such as color and size.
For attentive perception, it uses short term memory and it is selective. Attentive tasks convert premature image effects into a well-structured objects. Attentive perception is generally slower and often signifies aggregates of what is in the scene.
When designing visualization, the designers should take note of pros and cons of the human visual system and provide well-suited visuals to the viewers for easy analysis. Thus, in order to use the visual features effectively and not to produce visual interference effects masking information in a display, the visualization creators should be aware of the attentive tasks and the pre-attentive visual features like length, width, hue, intensity, lighting direction, and so on.
Data Foundation
The very primary step of visualization is the data to visualise. It is a must to explore and examine the characteristics of the data since it can be from many different kinds of sources and has a wide variety of attributes and features.
Data Types
Data can be differentiated into two main types: ordinal (numeric values) and nominal (non-numeric values). [2]
To be specific, ordinal values mean:
Binary values – those with only 0s and 1s
Discrete values – integer values from a very particular division
Continuous – real values
Nominal values are:
Categorical – values from list of possibilities
Ranked – categorical variables with significant ordering
Arbitrary – infinite range of values without significant ordering
Scale is another useful technique of sorting data variables since each graphical attribute from raw data possesses scale associated with it. There are three attributes of scale:
Ordering relation – ranked nominal variables and ordinal variables which can be ordered in some manner
Distance metric – all ordinal variables where the distances of different records can be calculated
Existence of absolute zero – variables with fixed lowest value
Data Pre-processing
In reality, real world data that is to be analyzed can be incomplete, noisy, incoherent and cumulative. Those raw data need to be transformed somehow into an understandable format and the process of its transformation is known as data pre-processing. Data pre-processing can greatly improve the quality of data visualization results. There are some different aspects of data pre-processing:
Metadata and statistical
Missing values and data cleansing
Normalization
Segmentation
Sampling and sub-setting
Dimension reduction
Mapping nominal dimensions to numbers
Aggregation and summarization
Smoothing and filtering
Raster to vector conversion
For more information about data pre-processing techniques refer to [2].
Visualization Techniques for Different Types of Data
Visualization techniques will be differed for different types of data since they comprise special characteristics. Main types of data and useful visualization methods for them will be discussed in this section.
Spatial Data
Spatial attributes identify data in 1,2 or multi dimension. Visualizing spatial data is defined as mapping spatial data to spatial attributes on the screen. [2]
Techniques of visualization of those data include histograms, linear probes, flow visualization, vector field visualization, slice plus isosurface, isosurface plus glyphs and so on.
Geospatial Data
Geospatial data or geographic information classifies geographic locations and boundaries in the real world. [8] They include coordinates and topology on earth. Examples of geospatial data consist of climate, environmental, economical and sociological and credit card payment locations.
Visualization methods of such data can be completed using dot maps, pixel maps, network maps, choropleth maps and cartograms. [2]
Multivariate Data
Multivariate data is lists or tables of data that arises from more than one variable. It normally doesn’t have an precise spatial attribute. [2]
Multivariate data can be visualized by point based techniques like scatter-plots and force based methods, line based techniques like graphs, parallel coordinates, andrews curves and radial axis techniques, and region based techniques which are bar charts, histograms and tabular displays. Combination of above techniques are also applied sometimes.
Trees, Graphs and Network
Bertin declares that trees, graphs and network visualization demonstrates the relationships of each data recorded, similarities among values and attributes, parent and child nodes, connectedness such as networks between countries around the world, shared classification and derivation. [9]
Space filling methods, non space filling methods, displaying arbitrary graphs and networks, and node link graphs are some of the methods for trees, graphs and networks visualization. maps, pixel maps, network maps, choropleth maps and cartograms. [2]
Text and Documents
By applying suitable visualization techniques, valuable information can be obtained from huge resources of information such as digital libraries, text files from your computer and billions of words in your thesis paper. Searching comparable patterns and outliers within the text or documents will be painful without visualization.
Tag clouds, word trees, text arcs and arc diagrams can be used for visualizing single documents. Visualization practice for collections of documents are self organizing maps, themescapes and document cards. [2]
Interaction Concepts & Techniques
John and his group clarify that interaction within data visualization is a helpful structure for transforming what the users see and how they perceive it. Interactions will transform visualization images to better and smooth transitions. Summary of interaction techniques are discussed as below. [10]
Navigation
It allows the users to adjust the camera’s position and scale the vision. Examples include panning, rotating and zooming.
Selection
Selection refers categorizing an object or collections of objects. To be precise, it grants the user to control the regions of interest. Highlighting, deleting and modifying are types of selection.
Filtering
The size of data mapped on the screen is reduced by filtering techniques by reducing or omitting dataset, dimensions or both.
Reconfiguring
It is to change the way analyzed data is mapped to visualization graphical attributes like reordering data layouts in order to provide a diverse way of viewing data.
Encoding
Users are permitted to control graphical attributes such as point size, line colors to discover different features of visualization.
Connecting
Connecting means linking different views or objects.
Abstracting and Elaborating
It is to modify the level of detail.
Hybrid
Hybrid defines combining the above techniques together.
Effective Visualization
In fact, visualizations implemented by the designers have larger risks of being ineffective than being effective. It is not very simple to build effective visualizations where the users satisfy as there are many chances of data being distorted and lost during the mapping process, or data presented is too confusing and complex for the users to interpret, and so on. A successful and effective visualization efficiently and accurately transmits the preferred information to the viewers. Therefore, the designers should take in consideration of what the targeted users really want to observe from the results so that they will be able to visualize effectively.
Intuitive Data Mappings
Ed H. Chi explains that it is essential to consider the importance of data semantics and the context of the user. [11] To avoid any misinterpretation, the designers should be able to predict the user’s expectations. Choosing data-to-graphics mappings that provides the user’s mental model will significantly support in interpretation. The designer should take note of the compatibility between scale of data and graphic attributes on the screen. Besides, they should utilize humans’ abilities to correlate position on the screen medium with position in real world.
Selecting and Adjusting Views
It is obvious that one view is hardly satisfactory to express all the information enclosed in the dataset. Expecting the view modifications which are most useful to the users is one of the major factors of developing an effective visualization. Common view operations are as follows. [2]
Scrolling and Zooming Operations
This operation comes in handy when the dataset is too huge to be presented as one whole at the resolution that the viewer wants.
Color Map Control
It allows the user to make changes of individual attribute colors or entire palette.
Mapping Control
Mapping control helps the viewers to toggle among different ways of visualizing the same data and to discover the distinct features which might be hidden.
Scale Control
The user can focus on specific data subsets by applying scale control where they can modify the range and distribution of values.
Information Density
The designer’s decision, to verify how much information to display, plays an important role for an effective visualization and representation. Alexandru [12] points out that if there is too little information to present, it is the best to display the results as text. Conversely, if the data has too much information to present, it might cause confusion, lose essential information within the data, and face with obscurities in interpretation. In such cases, the user should be permitted to disable or enable different components of the presentation.
Keys, Labels and Legends
Most of the visualizations are ineffective because they lack useful and supported information to aid them. [2] Keys, labels and legends are therefore very helpful. Examples include captions, mappings used, grid marks, units of axes, key for symbols, color bar and etc.
Using Color with Care
Color can add significant visual appeal to a visualization but can also significantly decrease the effectiveness of the communication process. [2] Usage of color is context dependent and the characteristics of dataset itself can influence how the colors are noticeable. The designers should not forget there might be some color blind users as well.
The Importance of Aesthetics
Visuals, with both informative and pleasing to the eye, are known as the best representations. If the visualization is aesthetically pleasant, it attracts the viewers to analyze it in greater details. [2] Some useful guidelines for attractive visualization designs are as below.
Focus
The user’s attention should be drawn towards the most vital part of the visualization.
Balance
Balancing the screen space is another aspect to take note of designing pleasing visualizations. The most important components should be placed in the center.
Simplicity
Representing too much information will confuse the viewers. The designers should get rid of features which can be removed without losing information wanted to pass on since it is the best to be as simple as possible.