Definitions, Design, and Parts
Bijan Parsia bparsia@cs.man.ac.uk Sean Bechhofer sean.bechhofer@manchester.ac.uk COMP60421 23 Nov. 2012
1 Monday, 26 November 2012
Definitions, Design, and Parts Bijan Parsia bparsia@cs.man.ac.uk - - PowerPoint PPT Presentation
Definitions, Design, and Parts Bijan Parsia bparsia@cs.man.ac.uk Sean Bechhofer sean.bechhofer@manchester.ac.uk COMP60421 23 Nov. 2012 Monday, 26 November 2012 1 Definition Oriented Development Define define Monday, 26 November 2012 2
1 Monday, 26 November 2012
2 Monday, 26 November 2012
3 Monday, 26 November 2012
– The languages – The services – The underlying logic – A bit about computation – A bit about how to model with them
– What are they good for? – (data integration, terminology development...)
– ...What is the marginal gain of using ontologies?
– Much more work needs to be done!
4
4 Monday, 26 November 2012
5
― Plato, Apology
5 Monday, 26 November 2012
has knowledge
Conceptualize
has Knowledge
Formalize Verbalize
6 Monday, 26 November 2012
has knowledge
Conceptualize
has Knowledge
F
m a l i z e Communicate
7 Monday, 26 November 2012
F
m a l i z e
has knowledge
Conceptualize
has Knowledge
Communicate I n f e r e n c e
8 Monday, 26 November 2012
– Local focus on what terms mean
– There are consequences to what we say – We can spot wrong links
– The reasoner can tell us about broken definitions
– The KR becomes “reactive”
9 Monday, 26 November 2012
– Aka taxonomies
– We must formulate the definitions – We must put terms “in their proper place” – We must assert every non-trivial “link” – We must check that these are the right links
– 100 terms ≈ 10,000 (1002) possible subsumptions!
– Depth adds complexity – Multiple inheritance adds significant complexity
10 Monday, 26 November 2012
11 Monday, 26 November 2012
h4p://www.cs.manchester.ac.uk/ugt/2011/COMP34512/slides/day7.pdf
12 Monday, 26 November 2012
13 Monday, 26 November 2012
14 Monday, 26 November 2012
15 Monday, 26 November 2012
* ¡h4p://bit.ly/zSxHpK
16 Monday, 26 November 2012
17 Monday, 26 November 2012
18 Monday, 26 November 2012
19 Monday, 26 November 2012
20
20 Monday, 26 November 2012
21
21 Monday, 26 November 2012
22
22 Monday, 26 November 2012
23
23 Monday, 26 November 2012
24
24 Monday, 26 November 2012
25
25 Monday, 26 November 2012
Email ¡Requests UniversiUes Research ¡ InsUtutes EVS ¡Partners Legacy ¡Data Bulk ¡Data ¡Imports Data ¡Archives Use ¡Cases
26 Monday, 26 November 2012
27
User ¡submits ¡a ¡Use ¡Case Domain ¡Expert ¡examines ¡use ¡case ¡to ¡ check: ¡
¡
The ¡collaboraUve ¡process ¡begins:
¡
proto-‑representaUon * ¡“Pseudo-‑TBox” ¡is ¡an ¡odd ¡artefact ¡of ¡the ¡rather ¡strange ¡DL ¡system ¡they ¡were ¡using ¡at ¡the ¡Ume
27 Monday, 26 November 2012
28
28 Monday, 26 November 2012
29 Monday, 26 November 2012
30 Monday, 26 November 2012
31 Monday, 26 November 2012
32 Monday, 26 November 2012
33 Monday, 26 November 2012
34
34 Monday, 26 November 2012
35
35 Monday, 26 November 2012
36
36 Monday, 26 November 2012
37
37 Monday, 26 November 2012
38
38 Monday, 26 November 2012
39
ip://ip1.nci.nih.gov/pub/cacore/EVS/ThesaurusSemanUcs/TBox.png
39 Monday, 26 November 2012
40
40 Monday, 26 November 2012
Ontology ¡Designers ¡rely ¡heavily ¡on ¡the ¡definiUons ¡and ¡comments ¡provided ¡ by ¡the ¡users ¡to ¡create ¡the ¡OWL ¡representaUons ¡of ¡the ¡definiUons
41
41 Monday, 26 November 2012
42
42 Monday, 26 November 2012
43
43 Monday, 26 November 2012
44
44 Monday, 26 November 2012
45 Monday, 26 November 2012
– When ¡to ¡stop – Task – Purpose – Output ¡-‑ ¡terminology
– Collect ¡terms – Organize ¡terms – Produce ¡informal ¡concepts ¡(proto-‑representaUon)
– Normalizing ¡terms ¡(e.g., ¡“symmetry ¡or ¡symmetric”?) – Hierarchy ¡(and ¡other ¡direct ¡relaUons ¡between ¡terms) ¡ – Categorizing ¡terms ¡(e.g., ¡as ¡modifiers ¡or ¡self-‑standing) ¡ – Constraining ¡and ¡defining ¡terms – Formalize ¡Knowledge
46
46 Monday, 26 November 2012
47
47 Monday, 26 November 2012
– No one hierarchy
domain expert about the hierarchy structure. Not infrequently in biomedicine, there is no canonical determination of a concept’s correct tree position. For example, meningococcal meningitis may be classified correctly as both a disease of the central nervous systems and a bacterial disease.” —Modeling a description logic vocabulary for cancer research
– Hierarchies aren’t neutral!
rationally map out a hierarchy of ideas...This presumes that there is a "correct" way of categorizing ideas, and that reasonable people, given enough time and incentive, can agree on the proper means for building a hierarchy. Nothing could be farther from the truth. Any hierarchy of ideas necessarily implies the importance of some axes over others.” —
Metacrap: Putting the torch to seven straw-men of the meta-utopia
48 Monday, 26 November 2012
Mary Van Rensselaer Buell (1893-1969) http://www.flickr.com/photos/smithsonian/3322785642/
http://www.well.com/~doctorow/metacrap.htm#2.5
49 Monday, 26 November 2012
50 Monday, 26 November 2012
– accommodate multiple compatible perspectives – unless our definitions conflict
– no silver bullet!
– If we can avoid pointless battles, yay! – Real disagreement is important!
51 Monday, 26 November 2012
52 Monday, 26 November 2012
http://www.cs.man.ac.uk/~rector/presentations/Reasoning-web-rector-GALEN-2006.ppt
53 Monday, 26 November 2012
collision with pedal cycle, person on outside of vehicle, nontraffjc accident, while working for income
highway, while engaged in sports activity
resting, sleeping, eating or engaging in other vital activities
http://www.cs.man.ac.uk/~rector/presentations/Reasoning-web-rector-GALEN-2006.ppt
54 Monday, 26 November 2012
– We are using opaque names – Thus every combination needs a new name – Thus must trade off
55
*http://online.wsj.com/article/SB10001424053111904103404576560742746021106.htm
55 Monday, 26 November 2012
– lots and lots and lots – how do we manage them?
– fundamental
– artificial (or “technical”)
56 Edwin Smith Papyrus
http://www.flickr.com/photos/78572993@N00/2226696853/
56 Monday, 26 November 2012
– Pedestrian / cycle / motorbike / car / HGV / train / unpowered vehicle / a tree / other
– Driving / passenger / cyclist / getting in / other
– resting / at work / sporting / at leisure / other
– In traffic / not in traffic
57
V12.24 Pedal cyclist injured in collision with two- or three-wheeled motor vehicle, unspecified pedal cyclist, nontraffic accident, while resting, sleeping, eating or engaging in other vital activities
Slide from: http://www.cs.man.ac.uk/~rector/presentations/snomed-rector-history-and-future-of-terminology.ppt
57 Monday, 26 November 2012
– from names to expressions – from few names we induce many expressions
– position driven by definition
– Terms have different grammatical roles
– We have to use logicy stuff
58
58 Monday, 26 November 2012
59
59 Monday, 26 November 2012
– No code without prior agreement
– Code drift
– Though devtime definitions can help!
– Core set of of terms
– A code is any valid expression
– Many nonsensical codes! – Requires reasoning at “runtime”
60
60 Monday, 26 November 2012
– Nuts and nut allergy
– No PecanAllergy, HazelnutAllergy, etc. – Why not?
– Very simple – We have a patient with an almond allergy
– Allergy?
61
61 Monday, 26 November 2012
– Nuts and nut allergy
– No PecanAllergy, HazelnutAllergy, etc. – Why not?
– Very simple – We have a patient with an almond allergy
– Allergy? – Nut Allergy?
62
62 Monday, 26 November 2012
– Nuts and nut allergy
– No PecanAllergy, HazelnutAllergy, etc. – Why not?
– Very simple – We have a patient with an almond allergy
– Allergy? – Nut Allergy? – Other?
63
63 Monday, 26 November 2012
– The fact of (divergent) choice – The uncontrolled bits
64
64 Monday, 26 November 2012
65
65 Monday, 26 November 2012
has knowledge Conceptualize
has 66
http://www.flickr.com/photos/seattlemunicipalarchives/4058808950/
66 Monday, 26 November 2012
F
m a l i z e
has knowledge
Conceptualize
has
C
m u n i c a t e I n f e r e n c e
67
http://www.flickr.com/photos/seattlemunicipalarchives/4058808950/
W r i t e e x p r e s s i
s
67 Monday, 26 November 2012
– Post-coordination ont to develop a pre-coordinated vocab
68
68 Monday, 26 November 2012
69
69 Monday, 26 November 2012
– E.g., Runtime has higher performance demands – E.g., Development time has higher correctness demands
– We can look at definition structure
70
70 Monday, 26 November 2012
71 Monday, 26 November 2012
– Depends on the effort for writing definitions
– “The true comparative cost of manual versus computed crafting and curating of classifications is not known. One unpublished report suggests that the semantic approach is less costly than continued manual maintenance, even for very small domains of only a few hundred concepts [Zantra 2004]: the manual addition of 20 new rubrics to the 2000-term musculoskeletal chapter of ICPM-DE required 6 full day meetings involving 10 specialists, and a total of more than 5 man months of effort spread over 8 months. The semantic-based computed approach required an initial one- time investment of 2.5 man months to represent the entire chapter of rubrics, after which the 20 new additions could be integrated very quickly.” – The marginal cost was negligible, and the total cost was half.
– I cordially invite you to provide some! – I’m happy to help
* http://www.opengalen.org/download/2004-Rogers-Methodology-and-ontological-schema-for-medical-terminology-(MD-Thesis).pdf
72 Monday, 26 November 2012
73 Monday, 26 November 2012
Expressivity (Representational Adequacy) Usability (Weak Cognitive Adequacy vs. Cognitive Complexity) Computability (vs. Computational and Implementational Complexity)
74 Monday, 26 November 2012
– What we can say (or say easily, or naturally, or...) – If suitable to our needs, a formalism (or KR) is representationally adequate
– How hard is it (in terms of resources) to work with? – Related: Implementational Complexity
– Focus on Weak Cognitive Adequacy i.e., Usability
75 Monday, 26 November 2012
– Indeed, we can view them as a (v. general) KR – The design considerations are similar
– FOL (at least) is well understood – FOL (at least) is v. expressive
– Usability.... – Computability...
76 Monday, 26 November 2012
– What expressivity do you need? – What's your core service? – What are the key services? – Are you interactive or not? – What's the scale you need to deal with?
– What do you know about implementation?
– Many surface syntax issues – Non logical aspects of the language – Cognitive complexity
77 Monday, 26 November 2012
– (Sometimes known: “theorem proving”) – (Typically: Deduction!) – Two basic flavors
– “Proof assistants” – A human user interacts with the program to generate proofs – System provides verification of steps, suggested next steps, etc. – “Mixed initiative”
– Set up initial conditions
– Default questions
– Program does the rest!
78 Monday, 26 November 2012
– Prove or verify math theorems – Verify algorithms – Things which are v. hard to prove
– Some math theorem proving – Whenever your concern is for the answer than the process
79 Monday, 26 November 2012
– Human intervention is required to supply proof steps
– E.g., a dialogue driven expert system – Consider database based systems – Consider Protege!
80 Monday, 26 November 2012
– Satisfiability/Consistency Checking
– Entailment
– Classification (atomic subsumptions) » Atomic class satisfiability – Instantiation
– More outré
– Verification, reaction, query
81 Monday, 26 November 2012
– User does not need to formulate the questions – Easy invocation of the questions – Built in coverage
– Given {A, B, C}, » check A SubClassOf: B; A SubClassOf: C, B SubClassOf:C...
– Formulation takes practice and inclination
– Users write to the default queries
– Reasoners get overtuned to default questions
Usability82 Monday, 26 November 2012
– Of a formalism?
– By tests! – But how to generate test input?
– We build a model of aspects of the computation
– How time/space varies as input size varies » Esp. “at the limit”
– We explore key cases
83
83 Monday, 26 November 2012
– For us, how about “number of axioms”
– Let’s stick with time
– Worst case – Best case – Average case – “Typical” case
– E.g., satisfiability
84
84 Monday, 26 November 2012
– I.e., yes/no questions
– (We can reduce many general problems to decision problems)
– parsing is a process which goes from strings to trees – recognising is a process which goes from strings to Yes/No
– Always many ways for a given problem – Some ways are better than others!
– They are related! – CoP is the complexity of the best possible algorithm
85
85 Monday, 26 November 2012
– For any (finite) string S and (finite) regular expression R
– (A*B|B*A)
– (Hint, regular expressions correspond to deterministic finite state machines.) – (With some work!) – Linear time, constant space
Computability (vs. Computational and Implementational Complexity)86 Monday, 26 November 2012
– Draw a line representing
– Do the same!
– for a single problem! (they are named, L, Q, and E :)) – what do we know about the WWC of this problem?
87
87 Monday, 26 November 2012
88
88 Monday, 26 November 2012
– Polynominal? I.e., “tractable”? – Does this rule out OWL immediately?
– Not necessarily! – Constants can matter – Worst case != Every case
– For a given input (or set of inputs)
– for our context’s value of “reasonable”
89
89 Monday, 26 November 2012
– for the same problem – What’s the WWC complexity of that problem?
– plot a set of inputs for each algo s.t. – we would prefer using E to Q and Q to L
90
90 Monday, 26 November 2012
– How often do you hit the worst case?
– What’s the best case?
– From a complexity POV!? – Tends to correspond with higher expressivity
91 Monday, 26 November 2012
Expressivity (Representational Adequacy) Usability (Weak Cognitive Adequacy vs. Cognitive Complexity) Computability (vs. Computational and Implementational Complexity)
92
92 Monday, 26 November 2012
– then the difficulty of working with the artifact goes up – consider different time scales!
– At least as the number of times it must be solved
93
93 Monday, 26 November 2012