G raded by Tarique and me Available during Tariques OH Proposals themselves Weve been very critical Take feedback seriously come meet us to discuss Dont stress about it Also annotated with the number of class reviews youve missed ID: 638764
Download Presentation The PPT/PDF document "Misc Project Proposals “" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
MiscSlide2
Project Proposals
“
G
raded” by
Tarique
and me
Available during
Tarique’s
OH
Proposals themselves
We’ve been very critical
Take feedback seriously: come meet us to discuss
Don’t stress about it
Also annotated with the number of class reviews you’ve missedSlide3
How does Wrangler make Potters Wheel ideas more “usable”?Slide4
How does Wrangler make Potters Wheel ideas more “usable”?
Suggestions?
Learned via ML
No dropdowns – painful interactions
Natural language description of transformations?
Semantic understanding?
Data quality meter
Additional semantics:
Position, aggregation, semantic roles, multiple selections, numbered selectionsSlide5
Language Comparison
Wrangler
Potters Wheel
Map
: Delete, Extract, Cut, Split
Divide,
Select, Drop, Split
Lookup and Joins
?
Reshape:
Fold, Unfold
Fold, Unfold
Positional:
Fill (copy)
and Lag (shift)
Copy
Sorting,
aggregation, key generation, schema modification
Types and semantics
Add, Merge?Slide6
Mechanisms for input
What are the various mechanisms for user specification of transformations?Slide7
Mechanisms for input
Direct manipulation
Automatic suggestion
Menu-based suggestion
Manual editing of transforms
Why are each of these interesting or important?Slide8
Parameter Identification
How can we improve identification of parameters of transformations within Wrangler?Slide9
Parameter Identification
How can we improve identification of parameters of transformations within Wrangler?
A 5 token window: labels of the form number, word, lowercase, uppercase, whitespace
More general, more semantically relevantSlide10
Criteria for ranking
What are the criteria that Wrangler uses for ranking transformations?Slide11
Criteria for ranking
What are the criteria that Wrangler uses for ranking transformations?
Explicit interaction
Specification difficulty
Frequency in corpus
Simplicity
Diversity
What else is missing?Slide12
Criteria for ranking
What are the criteria that Wrangler uses for ranking transformations?
Explicit interaction
Specification difficulty
Frequency in corpus
Simplicity
Diversity
What else is missing? Semantic knowledge, how much this improves the data quality, user preferences, …