CREATING KNOWLEDGE MAPS USING MEMORY ISLAND

Get Complete Project Material File(s) Now! »

Ontology and Ontology Visualization

Ontology is usually referred to as a formal and explicit description of concepts (classes) in a domain of discourse [29]. It contains the objects, concepts and other entities that are presumed to exist in some areas of interest and the relations that exist between them [30-32]. There are many mathematical definitions of ontology, such as those by Amann and Fundulaki [33] that can help in understanding how ontology can be processed by programs, and already wildly applied in the domain of Database.
Ontologies are useful to effectively present knowledge. The main reason ontologies reach outside the AI domain is their ability to support semantic linking, user interaction and visualization. For example, with the power of InPhO ontology[34], we can generate an insightful island for a large encyclopedia. Ontology enables many complex semantic relationships, associations, and interactions in a knowledge system to be formalized for processing by machines, which provides multiple ways of presenting or operating on the same set of data. For this reason, ontology visualization has attracted much interest with many research projects developing and testing methods, trying to find the best way of visualizing ontologies in order to achieve favorable outcomes for end-users.
The ontology visualization is not an easy task, because ontology is a data type richer than a tree structure dataset. The complexity of ontology involved including a hierarchy of concepts, concept attributes, concept relationships, and relationship roles. This is further complicated when concepts have thousands of instances attached to the concepts.
This problem is usually addressed in ontology visualization by reducing ontologies to an approximation of a hierarchical structure (tree-structure) that constitutes what is sometimes termed as “skeleton”. Usually, this skeleton gives a useful approximation of the ontology. However, low levels of user satisfaction in relation to the support of ontology visualization and exploration provided by current ontology visualizing tools [35].
In 2007, Katifori et al. [35] reviewed the existing works( published before July 2006) on ontology(taxonomy) visualization. They presented the techniques and methods and categorized their characteristics and features in order to assist method selection and promote future research in the area of ontology visualization. Besides ontology visualization techniques, they also included some tree or network (graph) visualization techniques, which are not created specifically for ontologies in their survey. They categorized existing techniques into the following six categories: Indented list: The windows explorer-like (file-explorer metaphor) tree view of the ontology (taxonomy), for example, the Protégé Class Browser [18]. See section 2.4.1 for details. Node-link diagram: represents taxonomy of ontologies as a set of interconnected nodes, see section 2.4.2.2. Normally these node-link diagrams allow their users to expand and retract nodes and their subtrees, in order to adjust the detail of the information shown and avoid display clutter (nodes’ overlaps). Zoomable: These techniques allow the user to zoom-in to the child nodes in order to enlarge them, making them readable in the viewing level. Grokker [36] is an example of this group.

Cartography and Knowledge Cartography

Many cartographic principle and technique have been introduced to create the knowledge map; these cartographic InfoVis approaches are also called Map-like visualizations, as they created a knowledge map as the visual representation of the dataset. In this section, we briefly review the basic concepts of Cartography, the Knowledge Cartographic and the existing map-like visualization techniques.

Cartography

Cartography, as its name, is the study on making maps. It is combining science, aesthetics, and technique. The main topic of cartography is how reality can be effectively modeled.
The fundamental problems of traditional cartography are to: Set the map’s agenda and select traits of the object to be mapped. This is the concern of map editing. Traits may be physical, such as roads or landmasses, or may be abstract, such as the (political) boundaries. Map projections: Represent the terrain of the mapped object on flat media (a 2D plane). Map generalization: The concern of generalization, how to make the map. o Eliminate characteristics of the mapped object that are not relevant to the map’s purpose. o Reduce the complexity of the characteristics that will be mapped. Map design: Orchestrate the elements of the map to convey best its message to its audience.

Knowledge Cartography

In recent years, some cartographers and researchers in InfoVis have tried to extend cartographic techniques to InfoVis for non-geographic information. These works are divided into two: Knowledge Cartography and cartographic/geographic visualization. Similar to the difference between Infographics (specific, handcrafted) and visualization (general, automatic)7, Knowledge Cartography is the discipline about mapping intellectual landscapes. It focus on how manually make an interactive, hyper-textual map for a knowledge, with one’s own understanding, and facilitates the communication process. Okada et al. wrote a book on the topic of knowledge cartography’s approaches [43].
Some outstanding maps have been created for a long time, such as Leonardo da Vinci’s Mappamundi (Figure 2.10) for geographic knowledge. This map has many exceptional properties, such as the earliest map showing that America is not connected to Asia. Recently, Marco Quaggiotto has proposed a knowledge cartographic tool : Knowledge atlas(an example shown in Figure 2.11) [44] [45] to help the knowledge experts manually craft their knowledge map.

Treemapping Approaches

Tree mapping approaches apply the space-filling algorithms for creating maps. They display hierarchical data as a map of nested regions (rectangles or non- rectangular regions). For example, the famous classic Tree-Map uses a space-filling algorithm to create the maps. In the beginning of 90s, the existing tree-drawing algorithms have problem for display a large tree-structure in a limited display space. This type of space filling algorithms also been considered as tiling algorithms, which try to fill the display space. The first treemapping approach was proposed by Johnson and Shneiderman [3]. They proposed a space-filling algorithm inspired by the idea of mosaic for tree-structure data. As this visualization technique creates a map (e.g., Figure 2.12) for the hierarchical data set (tree-structure dataset), Johnson and Shneiderman named their InfoVis technique as Treemaps.
Then there are many other space-filling (tiling) algorithms proposed after the work of Johnson and Shneiderman. All these algorithms tried to create a map by using nested rectangular regions or non-rectangular regions. In recent time, in 2013, Auber et al. [46] used the geometry of Goseper Curve(non-rectangular regions) to create Goseper maps. In 2014, Duarte et al. proposed their Nmap (Neighborhood Map) space-filling algorithm [47]. This Treemapping technique tried to keep the distance-similarity metaphor (between the concepts in the hierarchy) in its result Map. These approaches were widely applied in many domain, such as using in Disk space visualization tools for different operator systems. For example, in Figure 2.13, WinDirStat software applied a Tree Map for graphically displaying the amount of space used by files on a disk partition. With this interactive tree map, the end users can easily achieve the tasks of space managements of their disk. However, with these tilling algorithms, the relations between the concepts in the hierarchical knowledge become less evident.

Spatialization (Layout) and Tree drawing algorithms

Graph drawing algorithms create the spatialization of Graph datasets, the tree drawing algorithms are the algorithms that create the visual representations for a given hierarchical structure. Tree drawing is a specific part of the Graph Drawing. The Graph drawing algorithms are also called layout algorithm, as the output of these algorithms are the proxy elements distribution (layout). We can automatically generate a geographic representation by directly applying those algorithms for the given relational information.
In 2013, Rusu [48] summarized in detail the different tree drawing algorithms. To build a node-link graph visualization, one of the layout algorithms in Tree Drawing or Graph Drawing can be applied to create the 2D or multidimensional representation for the concepts or clusters. For example, the BubbleSets [49] create for Sets visualization, it build a map representation by drawing on the existing spatialization (tree or network spatial layout) use either the traditional convex hull or implicit surfaces(draw contiguous contours around nodes). Hong et al. summarized the classic layout algorithms for visualizations in [50]. Skupin and Fabrikant [51] summarized the existing spatialization for visualizing non-cartographic data.

Map‐like Visualization Approaches

Beside the Tree Map, the GMap algorithm[52] enclose group members with map metaphors(countries, seas and lakes) and the Self-Organizing Map(SOM) [53] approaches based on the clustering technique to build the 2D distribution for underlying data, are the most popular map-like visualization. One usefulness of this map-like visualization is that their result map can be used as basic maps to create many visualizations. For example, in the works of « Maps of Computer Science (Mocs)[54] » (an example is shown in Figure 2.15), Fried et al. used the GMap algorithm to generate a map and then overlap it with a heatmaps to create a map of computer science from different database.

Table of contents :

PART I. BACKGROUND AND LITERATURE REVIEW
CHAPTER 1 GENERAL INTRODUCTION
1.1 Overview and Research Context
1.2 Industrial Context
1.3 Contributions and Related Research Areas
1.4 Organization
CHAPTER 2 REVIEW OF LITERATURE
2.1 Information Visualization and knowledge visualization
2.2 InfoVis Toolkit and InfoVis Tools
2.3 Design and Evaluation Visualization Systems
2.4 Tree and Hierarchic Knowledge Visualizations
2.4.1 Tree, Hierarchic knowledge and Skeleton
2.4.2 Classical Tree Visualization Techniques
2.5 Ontology and Ontology Visualization
2.6 Knowledge Maps
2.7 Cartography and Knowledge Cartography
2.7.1 Cartography
2.7.2 Knowledge Cartography
2.8 Cartographic Visualization
2.8.1 Treemapping Approaches
2.8.2 Spatialization (Layout) and Tree drawing algorithms
2.8.3 Map‐like Visualization Approaches
2.9 Schematization
2.10 Spatial Cognition
PART II. MEMORY ISLAND TECHNIQUE
CHAPTER 3 MEMORY ISLAND IDEA AND DESIGN METHODOLOGY
3.1 The arts of Memory technique and the notion of Memory Islands
3.2 The Objectives of Memory Island technique
3.2.1 Recognition (truthful)
3.2.2 Discovery (functional and beautiful)
3.2.3 Surprise (insightful and enlightening)
3.3 The idea of Memory Island
3.4 Map (Landscape) Metaphor
3.5 Why do we choose 2D traditional map representation
3.6 Geographic Metaphors and Cartographic Means
3.6.1 Proportion Metaphor
3.6.2 Distance/centrality Metaphor
3.6.3 Font Attributes and Point Attributes
3.6.4 Colors
3.6.5 Paths and Gaps
3.7 Memory Island Prototype Algorithm
CHAPTER 4 HIERARCHICAL REORGANIZATION BY SEMANTIC SIMILARITY
4.1.1 Related works
4.1.2 Hierarchical Reorganization by Semantic Similarity
CHAPTER 5 VISUALIZING HIERARCHICAL DATA AS ISLANDS
5.1 Island Generation
5.1.1 Polyle II Algorithm
5.1.2 Memory Island Algorithm
5.1.3 Discussion on Memory Island Algorithm
5.2 Reshaping the Resulting Island
CHAPTER 6 LABELLING AND MAP GENERATION
6.1 Related Works
6.1.1 PFLP Problem
6.1.2 PFLP Algorithms
6.1.3 Google Map Mechanism
6.2 Labeling and Map Generation in Memory Island
6.3 Discussion
6.3.1 Improvement of the Label placement algorithm
6.3.2 Apply Area‐features label placement algorithm
CHAPTER 7 MEMORY ISLAND INTERFACE
7.1 Overview+Detail interactive interface
7.1.1 Map‐like Focus + Context and Element highlighting technique:
7.1.2 Map interactive functions
7.1.3 Design of Memory Island Interface
7.2 Discussion
PART III. IMPLEMENTATION AND APPLICATIONS
CHAPTER 8 MEMORY ISLAND IMPLEMENTATION
8.1 Memory Island Application’s sub‐system components
8.1.1 Knowledge Extraction
8.1.2 Island Generation
8.1.3 Label Placement and Map Generation
8.1.4 Memory Island Interface Generation
8.2 The run‐time of Memory Island Application
8.3 Discussion
CHAPTER 9 CREATING KNOWLEDGE MAPS USING MEMORY ISLAND
9.1.1 Text and Document Data
9.1.2 Ontologies
9.1.3 Hierarchical Dataset
9.1.4 Project OBVIL – Digital Humanities Dataset
9.2 Discussion
PART IV. VALIDATION AND EVALUATION OF MEMORY ISLAND
CHAPTER 10 VALIDATING MEMORY ISLAND
10.1 Visualization mantras
10.2 User Study
10.2.1 Methods and Visualization Set‐up
10.2.2 Subjects
10.2.3 Procedure
10.2.4 Result
10.3 Discussion
CHAPTER 11 PSYCHOLOGICAL EVALUATION
11.1 Related work
11.1.1 Users’ requirements analysis
11.1.2 Evaluations of Ontology Visualization tools
11.1.3 Evaluations of Tree Visualization tools
11.1.4 Evaluations of Map‐like Visualization tools
11.2 Psychological experimental protocol
11.2.1 Ontology Browsing Task
11.2.2 Ontology Understanding Task
11.2.3 Ontology Remembering Task
11.2.4 The Subjective Task
11.2.5 Suggested Evaluation Procedure
11.3 A Preliminary Users Study
11.3.1 Methods and Experiment Set‐up
11.3.2 Subjects
11.3.3 Procedure
11.3.4 Result
11.4 Discussion with the past evaluation experiment
PART VI. CONCLUSION AND PERSPECTIVES
CHAPTER 12 CONCLUSIONS
CHAPTER 13 FUTURE WORKS
13.1 Develop the Visualization Tools for Digital Humanities
13.2 Improve the Usage of space
13.3 Evaluation of the long‐term memorization
13.4 User Interface and interactive functions
13.5 Integrate Memory Islands to platforms of DH and DL
REFERENCES

GET THE COMPLETE PROJECT