Enhancing Worker Digital Experience: A Hands-on Workshop for Partners
Ldml application - public
1. Lightweight Data Markup Language
Applications
Sayandeep Khan
Drakoon Aerospace
Invention Report
Public Release
May 27 2012
2. Contents
→LDML
⬔ Review
⬔ Example
→Multidimensional Data
⬔ Dead Ends
⬔ Translation : Description guided action
→Multiple Data Source
⬔ Coupling
→Metadata
⬔ Basics
⬔ Application : Machine guided investigation
3. LDML Review
Lightweight Data Markup Language
UTF-8
Multiple Nodes in One Document
Strict Inter-sentence Relation
A graph can be formed
4. LDML Example
Example Data Corresponding Tree
River R R P
Channel Size Sc Ef
Carriage Capacity Cc
Vg
Actual Carriage Ac Sc
Slit Capacity Ss Ac
Slit Load SL
Vegetation Vg CC
SL
Rainfall P Ss
5. LDML Example
⬔ In the last Model:
Orange : Attribute Relation, i.e. R>(Sc, Ac, Ss, Cc)
Yellow : Detemination Relation, i.e. Sc > Cc>Cs
Notice Sc uniquely determines the other two. It is possible to
arrive at Cc and Cs from just Sc, along a single branch of the
tree, whose root is known. This is dead-end, and can be
removed. Just save Sc.
⬔ Actuall Carriage however depends on Rainfall, and
External factors Ef. Ef may be unknown. Therefore this is not
Dead end
⬔ Remove Deadends to improve storage!!!
6. Description Guided Action
⬔ Take the statements: R>(Sc, Ac, Ss, Cc,SL) and Sc >Cc> Ss
⬔ Parser must look for definitions of each attribute
→ Deduction relation, Dead end : Calculate, Stop
→ Deduction relation without dead end: Search other
variables. Alert if missing.
⬔ Example: Sc, Ac, Ss, are defined. Sc >Cc, therefore can
calculate Cc. Sc >Cc>Ss , therefore, delete Ss .(Ac ,Vg,Ef)>SL.
Therefore either define SL or define (Vg ,Ef)
7. Multiple Data Sources
⬔ The data on river and vegetation are from
different sources (surveys).
⬔ Connected by relations
⬔ Clustering of Datasets
8. Metadata - Basics
⬔ Describes a Dataset.
▣ Assume Dataset River
→ Channel size
→ Carriage capacity
→ Slit capacity
→ Actuall Carriage
→ Slit Load
⬔ Metadata is :
R>(Sc,Cc,Ss,Ac,SL)
Sc>Cc>Ss
⬔ Unique relations, regardless to variable name
9. Metadata Comparison
⬔ Graphs
→ Each element a node
→ Each relation an edge
→ Different relations have different weighted value
⬔ Equivalent Datasets
→ Isomorphic graphs
⬔ Comparison by isomorphism test
10. Metadata Comparison: example
q
A C
s
r
B
D
t
E
u
F p
Dataset 1
Dataset 2
⬔ Orange : Attributes, value: a, Yellow: Deduction,
value: b. Both graphs are isomorphic → same dataset, A
= p, etc
11. Conclusion
⬔ Using relation between attributes:
→ Remove unnecessary storage
→ Find if information missing
→ Compare datasets (explianation of variable names
not needed)
→ Couple multiple datasets
→ Calculate missing information
⬔ In future: any algebric procedure applies
⬔ LDML : Capable of scientific manipulation of Data