Presented by Mikhail Khludnev, eCommerce Search Platform, Grid Dynamics This talk describes our experience in eCommerce Search: challenges which we’ve faced and the chosen approaches. It’s not indented to be a full description of implementation, because too many details need to be touched. This talk is more like problem statement and general solutions description, which have a number of points for technical or even academic discussion. It’s focused on text search use-case, structures (or scoped) search is out of agenda as well as faceted navigation.

1. CONCEPT SEARCH FOR ECOMMERCE
WITH SOLR
Mikhail Khludnev
Principal Engineer,
eCommerce Search Team
mkhludnev@griddynamics.com
http://goo.gl/2OHk2

2. Agenda
● eCommerce Search is Special
● Subphrase Query Parser
● Staged Search
● Match Spotting

3. Part I
eCommerce Search is Special

5. Boolean
Retrieval
(+ -)
Vector
Space
Model
http://upload.wikimedia.
org/wikipedia/commons/f/ff/Vector_space_
model.jpg

6. Does she search this way?
http://tustinhairsalon.com/wp-
content/uploads/2011/02/Blonde_Hair_stylist_s
tylists_blondes_tustin_santa_ana_orange_cou
+"michael kors" type:handbag

7. http://tustinhairsalon.com/wp-
content/uploads/2011/02/Blonde_Hair_stylist_s
tylists_blondes_tustin_santa_ana_orange_cou
michael kors hand bag
That's how she searches!

8. Plain Text Documents
Else Jeans Skinny Jeans,
Colored Denim Dark Green-
Wash
In a dark green wash perfect
for fall, these Else Jeans skinny
jeans hit the colored-denim
trend right on the mark! Lyocell
cotton rayon polyester Lycra
Machine washable Imported
Low rise: approx. 8 inches
Skinny fit Skinny leg Zipper fly
with button closure 5-pocket
style Dark green wash, colored
denim Waistband with belt
loops Inseam: approx. 30-1/2
inches
Lorem ipsum dolor sit amet,
consectetur adipisicing elit, sed
do eiusmod tempor incididunt
ut labore et dolore magna
aliqua. Ut enim ad minim
veniam, quis nostrud
exercitation ullamco laboris nisi
ut aliquip ex ea commodo
consequat. Duis aute irure
dolor in reprehenderit in
voluptate velit esse cillum
dolore eu fugiat nulla pariatur.
Excepteur sint occaecat
cupidatat non proident, sunt in
culpa qui officia deserunt mollit
anim id est laborum.

9. Plain Text Documents
Else Jeans Skinny Jeans,
Colored Denim Dark Green-
Wash
In a dark green wash perfect
for fall, these Else Jeans skinny
jeans hit the colored-denim
trend right on the mark! Lyocell
cotton rayon polyester Lycra
Machine washable Imported
Low rise: approx. 8 inches
Skinny fit Skinny leg Zipper fly
with button closure 5-pocket
style Dark green wash, colored
denim Waistband with belt
loops Inseam: approx. 30-1/2
inches
Lorem ipsum dolor sit amet,
consectetur adipisicing elit, sed
do eiusmod tempor incididunt
ut labore et dolore magna
aliqua. Ut enim ad minim
veniam, quis nostrud
exercitation ullamco laboris nisi
ut aliquip ex ea commodo
consequat. Duis aute irure
dolor in reprehenderit in
voluptate velit esse cillum
dolore eu fugiat nulla pariatur.
Excepteur sint occaecat
cupidatat non proident, sunt in
culpa qui officia deserunt mollit
anim id est laborum.
approx velit

10. BRAND: Calvin Klein
GENDER: Women's
TYPE: Jeans
FIT: At Waist
OCCASION: Casual
LEG: Classic Straight
WEIGHT: Super Skinny
COLOR: Light Wash
Product Catalog Document

11. Polysemy Problem

12. Polysemy Problem
COLOR:"Pink" TYPE:"Sweater"
BRAND: "Tomas Pink" TYPE:"Sweater"
BRAND: "Pink Lotus" TYPE:"Sweater"

13. http://www.shopstyle.com/browse/sweaters?
fts=pink+sweater&fl=b7&fl=b28462&fl=b1830&fl=b15382&fl=b12130&fl=b737

14. http://www.shopstyle.com/browse/sweaters?
fts=pink+sweater&fl=b7&fl=b28462&fl=b1830&fl=b15382&fl=b12130&fl=b737

16. pink jeans by google
https://www.google.com/search?hl=ru&tbm=shop&q=pink+sweater&oq=pink+sweater&gs_l=products-cc.3..
0l7j0i5l3.4658.7244.0.7699.12.11.0.1.1.0.298.1153.5j4j1.10.0...0.0...1ac.1.
denzbOMlNDg#q=pink+sweater&hl=ru&tbm=shop&ei=oDmIUPXgAoqC4gSylYGACg&start=20&sa=N&num=20&

17. pink jeans by google
https://www.google.com/search?hl=ru&tbm=shop&q=pink+sweater&oq=pink+sweater&gs_l=products-cc.3..
0l7j0i5l3.4658.7244.0.7699.12.11.0.1.1.0.298.1153.5j4j1.10.0...0.0...1ac.1.
denzbOMlNDg#q=pink+sweater&hl=ru&tbm=shop&ei=oDmIUPXgAoqC4gSylYGACg&start=20&sa=N&num=20&

18. right pink jeans

19. right pink jeans

20. Polysemy Problem No.2
BRAND:"Pink Rose" TYPE:"Sweater"
COLOR:"Pink","Rose" TYPE:"Sweater"
sweater

21. pink rose jeans
problem
pink rose sweater

22. precise pink rose jeans
pink rose sweater

23. Ranking Can't Help

24. Ranking Can't Help

25. Precision
is a MUST!

26. http://tustinhairsalon.com/wp-
content/uploads/2011/02/Blonde_Hair_stylist_stylists_blondes_tustin_
santa_ana_orange_county.jpg
Think About the User

28. http://www.amazon.com/Silver-Jeans/e/2206256011

29. Wrap-Up
● Product Descriptions Differs to Plain Text
● No Query Syntax
● Polysemy Problem
● Ranking Can't Help When Precision is Low
eCommerce Search is Special!

30. Part II
Subphrase Query Parser

31. Don't tokenize at index time

32. black and decker toaster oven
"black"
"black and"
"black and decker"
... "and"
"and decker"
"and decker toaster"
... "decker"
"decker toaster"
"decker toaster oven"
"toaster oven"

33. black and decker toaster oven
"black"
"black and"
"black and decker"
... "and"
"and decker"
"and decker toaster"
... "decker"
"decker toaster"
"decker toaster oven"
..."toaster oven"
ID:
123
BRAND: "black and
decker"
TYPE: "toaster
oven"
COLOR: white

34. black and decker toaster oven
"black"
"black and"
"black and decker"
... "and"
"and decker"
"and decker toaster"
... "decker"
"toaster"
"decker toaster oven"
... "oven"
ID:
456
BRAND: "black and
decker"
TYPE: "oven"
COLOR: white

35. black and decker toaster oven
"black"
"black and"
"black and decker"
... "and"
"and decker"
"and decker toaster"
... "decker"
"decker toaster"
"decker toaster oven"
..."toaster oven"
ID:
789
BRAND: "decker
toaster"
TYPE: "toaster
oven"
COLOR: black

36. fee people sweater

38. Part III
Staged Search

39. Precision and Recall
Expected

40. ExpectedActual

41. ExpectedActual
×× OK
True
Positive
False
Negative
False
Positive

42. Precision =
Expected
Actual
Actual ∩ Expected
Actual

43. Expected
Actual
Recall =
Actual ∩ Expected
Expected

44. Expected
Actual
100% Recall

45. Expected
Actual
Tuning Trade-Off

46. Expected
Actual
Tuning Trade-Off

47. Expected
Actual
Tuning Trade-Off

48. Expected
Actual
Tuning Trade-Off

49. defaultOperator
AND OR
numFound=100 numFound=1000

50. defaultSearchField / queryFields
BRAND BRAND
TYPE
STYLE
text
numFound=100 numFound=1000

51. Stemming & Synonyms
text text_stemmed
numFound=100 numFound=1000

52. BRAND:(Silver AND Jeans)
BRAND:(Silver OR Jeans)
BRAND_stem:(Silver OR Jean)
OR
text_stem:(Silver OR Jean)
ORDer by DESCending Precision
numFound=100
numFound=1000

53. BRAND:(Silver AND Jeans)
BRAND:(Silver OR Jeans)
BRAND_stem:(Silver OR Jean)
OR
text_stem:(Silver OR Jean)
numFound>0
numFound>0

54. Searching Complexity
O(n log k)
It Depends™
num found
page size

55. Stages Ratios
50%
30%
20%
Compounds - Exact Match
Compounds - Omit Words
Text Match

56. http://tustinhairsalon.com/wp-
content/uploads/2011/02/Blonde_Hair_stylist_stylists_blondes_tus
Think About The User

57. Part IV
Match Spotting

58. Facet Counts

59. Explain Info
15.110098 = (MATCH) sum of:
11.197777 = (MATCH) weight(BRAND:Free People in 28807), product
of:
0.81810623 = queryWeight(BRAND:Free People), product of:
9.678479 = idf(docFreq=232, maxDocs=1368899)
0.08452839 = queryNorm
13.687436 = (MATCH) fieldWeight(BRAND:Free People in 28807), product of:
1.4142135 = tf(termFreq(BRAND:Free People)=2)
9.678479 = idf(docFreq=232, maxDocs=1368899)
1.0 = fieldNorm(field=BRAND, doc=28807)
3.912321 = (MATCH) weight(PRODUCT_TYPE:SWEATER in
28807), product of:
0.5750671 = queryWeight(PRODUCT_TYPE:SWEATER), product of:
6.8032427 = idf(docFreq=4130, maxDocs=1368899)
0.08452839 = queryNorm
6.8032427 = (MATCH) fieldWeight(PRODUCT_TYPE:SWEATER in 28807),
product of:
1.0 = tf(termFreq(PRODUCT_TYPE:SWEATER)=1)
6.8032427 = idf(docFreq=4130, maxDocs=1368899)
1.0 = fieldNorm(field=PRODUCT_TYPE, doc=28807)

60. Match
Spotting
Explain
Info
Facet
Counts

61. BRAND:"silver jeans" TYPE:"dress" STYLE:"white"
BRAND:"alfani" TYPE:"dress" STYLE:"silver","jeans"
BRAND:"chaloree" TYPE:"dress" STYLE:"silver"
BRAND:"style&co" TYPE:"jeans dress" STYLE:"silver"
BRAND:"silver jeans" TYPE:"dress" STYLE:"black"
BRAND:"silver jeans" TYPE:"dress" STYLE:"white"
BRAND:"silver jeans" TYPE:"jacket" STYLE: "black"
BRAND:"angie" TYPE:"dress" STYLE:"silver","jeans"
BRAND:"chaloree" TYPE:"jeans dress" STYLE:"silver"
BRAND:"silver jeans" TYPE:"dress" STYLE:"blue"
BRAND:"dotty" TYPE:"dress" STYLE:"silver","jeans"
BRAND:"chaloree" STYLE:"jeans" "dress"

62. BRAND:"silver jeans" TYPE:"dress" STYLE:"white"
BRAND:"alfani" TYPE:"dress" STYLE:"silver","jeans"
BRAND:"chaloree" TYPE:"dress" STYLE:"silver"
BRAND:"style&co" TYPE:"jeans dress" STYLE:"silver"
BRAND:"silver jeans" TYPE:"dress" STYLE:"black"
BRAND:"silver jeans" TYPE:"dress" STYLE:"white"
BRAND:"silver jeans" TYPE:"jacket" STYLE: "black"
BRAND:"angie" TYPE:"dress" STYLE:"silver","jeans"
BRAND:"chaloree" TYPE:"jeans dress" STYLE:"silver"
BRAND:"silver jeans" TYPE:"dress" STYLE:"blue"
BRAND:"dotty" TYPE:"dress" STYLE:"silver","jeans"
BRAND:"chaloree" STYLE:"jeans" "dress"
silver jeans dress

63. BRAND:"silver jeans" TYPE:"dress" STYLE:"white"
BRAND:"alfani" TYPE:"dress" STYLE:"silver","jeans"
BRAND:"chaloree" TYPE:"dress" STYLE:"silver"
BRAND:"style&co" TYPE:"jeans dress" STYLE:"silver"
BRAND:"silver jeans" TYPE:"dress" STYLE:"black"
BRAND:"silver jeans" TYPE:"dress" STYLE:"white"
BRAND:"silver jeans" TYPE:"jacket" STYLE: "black"
BRAND:"angie" TYPE:"dress" STYLE:"silver","jeans"
BRAND:"chaloree" TYPE:"jeans dress" STYLE:"silver"
BRAND:"silver jeans" TYPE:"dress" STYLE:"blue"
BRAND:"dotty" TYPE:"dress" STYLE:"silver","jeans"
BRAND:"chaloree" STYLE:"jeans" "dress"

64. BRAND:"silver jeans" TYPE:"dress"
TYPE:"dress" STYLE:"silver","jeans"
TYPE:"jeans dress" STYLE:"silver"
BRAND:"silver jeans" TYPE:"dress"
BRAND:"silver jeans" TYPE:"dress"
TYPE:"dress" STYLE:"silver","jeans"
TYPE:"jeans dress" STYLE:"silver"
BRAND:"silver jeans" TYPE:"dress"
TYPE:"dress" STYLE:"silver","jeans"

65. BRAND:"silver jeans" TYPE:"dress"
TYPE:"dress" STYLE:"silver","jeans"
TYPE:"jeans dress" STYLE:"silver"
BRAND:"silver jeans" TYPE:"dress"
BRAND:"silver jeans" TYPE:"dress"
TYPE:"dress" STYLE:"silver","jeans"
TYPE:"jeans dress" STYLE:"silver"
BRAND:"silver jeans" TYPE:"dress"
TYPE:"dress" STYLE:"silver","jeans"

66. BRAND:"silver jeans" TYPE:"dress" (4)
TYPE:"dress" STYLE:"silver","jeans"
TYPE:"jeans dress" STYLE:"silver"
TYPE:"dress" STYLE:"silver","jeans"
TYPE:"jeans dress" STYLE:"silver"
TYPE:"dress" STYLE:"silver","jeans"

67. BRAND:"silver jeans" TYPE:"dress" (4)
TYPE:"dress" STYLE:"silver","jeans"
TYPE:"jeans dress" STYLE:"silver"
TYPE:"dress" STYLE:"silver","jeans"
TYPE:"jeans dress" STYLE:"silver"
TYPE:"dress" STYLE:"silver","jeans"

68. BRAND:"silver jeans" TYPE:"dress" (4)
TYPE:"dress" STYLE:"silver","jeans" (3)
TYPE:"jeans dress" STYLE:"silver"
TYPE:"jeans dress" STYLE:"silver"

69. BRAND:"silver jeans" TYPE:"dress" (4)
TYPE:"dress" STYLE:"silver","jeans" (3)
TYPE:"jeans dress" STYLE:"silver" (2)

70. BRAND: "Silver Jeans" - (156)
COLOR: "Silver" TYPE: "Jeans" - (34)
found 190

71. http://www.amazon.com/Silver-Jeans/e/2206256011

72. BRAND: "Silver Jeans" - (156)
COLOR: "Silver" TYPE: "Jeans" - (34)
found 190

73. Wrap-Up
● eCommerce Search is Special
● Subphrase Query Parser
● Staged Search
● Match Spotting

74. CONTACT
Mikhail Khludnev
mkhludnev@griddynamics.com
http://goo.gl/2OHk2