%0 Thesis %1 ng00 %A Gary, G. K. C. %B Computer Science %C Manchester %D 2000 %I University of Manchester %K ontology information knowledge acquisition visualization %T Interactive Visualization Techniques for Ontology Development %U http://www.cs.toronto.edu/~nernst/papers/GaryNg-Thesis-InfoLens.pdf

@phdthesis{ng00, added-at = {2006-03-24T16:34:33.000+0100}, address = {Manchester}, author = {Gary, G. K. C.}, biburl = {https://www.bibsonomy.org/bibtex/2955fe7c90d373c2b3674ee34988de295/neilernst}, booktitle = {Computer Science}, citeulike-article-id = {121819}, comment = {Ng thesis ip.23 - dl description - simple types have no roles associated with them e,g car is atomic, and other concepts can be defined as a result of it - e,g, car accelration_rrate Fast is a defined sports car. p. 25 GRAIL defines which classes may be combined (role sanctioning) p. 27 dynamic creation of the taxonomy based on what sanctions there are p. 28 the advantages of a description logic are also its modelling isadvantages e.g what is actually going to be created when classified p.29 types of consistency - domain and modelling, where modelling is the agreement between teams of users p. 30 misplaced general roles (trauma has location Bone) is too restrictive but leads to unexpected classification how to do this? proactive creation and reacitive problem solving p33 thesis makes the point that the tool must support user sets of concepts to look at p. 36 look at the background references he mentions (some have been aplied to info p. 39 the relationships between the different components of an information visualization p. 42 high level task you want to use with your data (Yu) Sneiderman: overview first, zoom and filter, details on demand - informationeeking mantra, is this true for all users? I agree that tihis is true for, say, me attempting to understand how to use the wines ontology, but why would someone doing NCI work care about this?! this thesis presents a lot of data on why info Vis is usueful, but I don;'t think this justifies the solution to the problem. OK, there are these abilities, but show me how it applies to this tool. what about the idea that when you choose an inforvis for people, you are predetermining what they can see? - what is this correct visualization? sufficient - have a beard, If true, you are male, if not doesn't prove you aren't. necessary - you must have a penis to be male. (seems backwards to what is described in the thesis. Sufficiency indicates intrinsic properties while necessity indicates p. 55 the reference ontology and the domain model. RO is the system classes level. mentions the split of certain groups of concepts into functional categories. but what about intergroup relationships? p56 tthree tyes of GRAIL users (not more?) - modeller, application developer,end user - also new users - p. 57 GRAIL is a parsimonious modeling framework. MOdeling is like SENG. Thesis focused on modelers. p. 58 modeling related tasks:authoring, verifying, exploring This whole area is pretty important to read over. In general, I think DLhas this a a central problem and so for SWeb stuff willbe even more important. Since you never really see the ihierarcy until it is defined. These seem quite general but where do they comefrom, and how do we know they are exhaustive? p. 64 Browsing - navigate and explore We have four categories of functional requirements but this guy presents three usage models for knowledge engineering: explore, verify, author - question: where are they at now, 3 years later - have they solved these issues? p 69 table is quite useful - "provide a tool which wouldprovide easier access to sets of concepts, and facilitates their comparison during verification and exploration" Chapter 3 - lit review GKB-Editor has many similarities to space tree see CODE4 for node-link diagram see good iscussion of different visualization requirements at end of chapt 3 chapter 4 reject node link diagrams. He used nodes as both concepts and roles because multiple concpets could have more than one associated trole, and I guess he didn't want to lose the sense that roles were first order objects. (seems simplistic,) p. 113 however, a concept such as Hip had 174 roles?! hard to show hierarchy in the matrix version landscape view provides a good overview and abstraction of the underlying data, thi is an area where metrics would be very useful ("how many X are near Y") (opportunistic browsing) see paper p 123 for types of metrics constellations (p 128) seem quite interesting require ability to connect directly to protege model in class tab su barrier to entry is much lower. Do this and trY and show that it increases adoption}, description = {sdasda}, interhash = {ec4a5e9aad0a938148fd705526e913f6}, intrahash = {955fe7c90d373c2b3674ee34988de295}, keywords = {ontology information knowledge acquisition visualization}, priority = {0}, publisher = {University of Manchester}, timestamp = {2006-03-24T16:34:33.000+0100}, title = {Interactive {V}isualization {T}echniques for {O}ntology {D}evelopment}, url = {http://www.cs.toronto.edu/~nernst/papers/GaryNg-Thesis-InfoLens.pdf}, year = 2000 }