6. ¡ Competitive
Landscape
§ Who
are
your
competitors?
§ How
are
they
pricing
products?
§ What
other
products
do
they
carry?
¡ Scale
§ Products
§ Sites
§ Categories
Store
Product
Match
Matching
is
Central
to
answering
key
questions
in
retail
analytics
7. 1. Title
2. Image
URL
3. Price
4. Description
5. Tables
Challenges:
Scale,
Depth,
Diversity,
Change
DOM
Tree
Title
or
Not:
Binary
Classification
Class
Imbalance
8. DOM
Tree
HTML
Features
Visual
Features
Random
Forest
Model
9. 1. Title
2. Image
URL
3. Price
4. Description
5. Tables
Category
Taxonomy
Challenges:
Large
Taxonomy,
Lack
of
Training
Data,
Changes
in
Taxonomy
10. Linear
SVM
CNN
Ensemble
Breadcrumb
Mapping
Background
Knowledge
11. 1. Title
2. Image
URL
3. Price
4. Description
5. Tables
Challenges:
Large
number
of
attributes,
bad/missing
data,
variability
1. Brand
2. Size
3. Color
4. Packs
5. …
Schema
14. 1. Title
2. Image
URL
3. Price
4. Description
5. Tables
Challenges:
No
single
approach
works
well
1. Brand
2. Size
3. Color
4. Packs
5. …
Category
Enriched
Product
Record
16. Challenges:
Pairwise
Distance
Computation,
Match
at
a
Store
Constraint
Store
Product
Match
1. Pairwise
Distance
Computation
2. Constrained
Clustering
18. ¡ Constraint
Type
§ Must
Link
§ Cannot
Link
¡ Examples
§ UPC
§ MPN
§ Match
at
a
Store
Must
Link
Cannot
Link
May
Link
D(P, P0
)
Use
Constrained
Clustering
19. Parsing
Classification
Attribute
Extraction
Blocking
Match
Inference
HTML
Product
Record
Classified
Products
Attributes
Product
Groups
Matches
20. Reported
Actual
Correct
Correct
Actual
Reported
¡ Precision
§ Sample
and
Spot-‐check
¡ Recall
§ Hard
to
estimate
§ Rare
population
§ Manually
search
products
on
a
site
to
produce
blind
sets
Lack
of
Ground
Truth
is
the
biggest
road
block
Correct