Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Sign up for free
Menu
Search
Features
All features
Private URLs
Password Protection
Custom URLS
Scheduled publishing
Remove Branding
Restrict embedding
Deck Collections
Notes
Features
All features
Private URLs
Password Protection
Custom URLS
Scheduled publishing
Remove Branding
Restrict embedding
Deck Collections
Notes
Explore
Featured decks
Featured speakers
Programming
Technology
Storyboards
Explore
Featured decks
Featured speakers
Programming
Technology
Storyboards
Pricing
Search
Sign in
Sign up for free
User Intent Behaviour and Perceived Satisfactio...
Search
ohnabe
April 14, 2018
Research
780
0
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
User Intent Behaviour and Perceived Satisfaction in Product Search
WSDM2018 論文読み会
ohnabe
April 14, 2018
More Decks by ohnabe
See All by ohnabe
並列処理を頑張って、睡眠時間を確保する
ohnabe
0
98
Other Decks in Research
See All in Research
EIRによる不正端末のブロッキング 5G時代におけるデバイス識別と不正対策の進化
stellarcraft
0
130
PHTalks Bengaluru - SSRF When All Else Fails
dk999
0
1.1k
ふとした出会いで生まれたSkillが、 社内利用1位になるまで
mikimhk
21
24k
Source Code Diff Revolution
tsantalis
0
140
敵対生成プロンプト同時探索による内省型プロンプト最適化
kinoue_smarthr
0
410
2026年度 生成AI を活用した論文執筆ガイド/ワークショップ / 2026 Academic Year Guide to Writing Papers Using Generative AI - Workshop
ks91
PRO
0
230
Apache Gravitinoで実現する Icebergカタログ統合とアクセスの一元化
matsumooon
0
500
データサイエンティストの就労意識~2015 → 2026 一般(個人)会員アンケートより
datascientistsociety
PRO
0
780
多様なデータを許容し学習し続ける模倣学習 / Advanced Imitation Learning for VLA
prinlab
0
300
MM-OVSeg: Multimodal Optical–SAR Fusion for Open-Vocabulary Segmentation in Remote Sensing
satai
3
130
Fukui Shibiten 39 - AI Art
butchi
0
190
HackSick vol.7 LT資料【LLMアーキテクチャ入門・事前学習時の躓き所解説】 スパースなAttention・状態空間モデル
rikkabotan7
0
170
Featured
See All Featured
The Myth of the Modular Monolith - Day 2 Keynote - Rails World 2024
eileencodes
28
3.6k
What's in a price? How to price your products and services
michaelherold
247
13k
Avoiding the “Bad Training, Faster” Trap in the Age of AI
tmiket
0
230
Abbi's Birthday
coloredviolet
3
9.9k
How to train your dragon (web standard)
notwaldorf
97
6.8k
Keith and Marios Guide to Fast Websites
keithpitt
413
23k
sira's awesome portfolio website redesign presentation
elsirapls
0
410
Building Applications with DynamoDB
mza
96
7.2k
Pawsitive SEO: Lessons from My Dog (and Many Mistakes) on Thriving as a Consultant in the Age of AI
davidcarrasco
0
240
Helping Users Find Their Own Way: Creating Modern Search Experiences
danielanewman
31
3.4k
Ten Tips & Tricks for a 🌱 transition
stuffmc
0
210
GraphQLの誤解/rethinking-graphql
sonatard
75
12k
Transcript
User Intent, Behavior, and Perceived Satisfaction in Product Search Takafumi
Watanabe(@ohnabe) WSDM 2018จಡΈձˏΫοΫύου 2018/04/14
ࣗݾհ • ໊લɿᬒจɹ@ohnabe • ॴଐɿIdein Inc • ࣄɿػցֶशͷΤϯδχΞͱͯ͠ɺ৭ΜͳϞσ ϧΛ࡞ͬͨΓͯ͠ɺࠓը૾ೝࣝͷϞσϧΛϝΠ ϯͰ࡞͍ͬͯΔͷͰɺD◦◦p
Learning ͱ͓͖ ߹͍͍ͯ͠·͢ɻ
ຊจͷ֓ཁ • Ϣʔβ͕ΦϯϥΠϯγϣοϐϯάΛ͢ΔࡍʹͲ ͏͍͏ҙਤΛ͍࣋ͬͯΔ͔Λఆٛ • ͦͷҙਤ͕γϣοϐϯάதͷϢʔβͷߦಈʹͲ ͷΑ͏ͳҧ͍ͱͯ͠ݱΕΔ͔Λߟ • ϢʔβͷߦಈΛجʹγϣοϐϯάதͷຬΛ ਪఆ͢ΔϞσϧΛߏங͠ɺ࣮ݧʹΑΓϞσϧͷ
ධՁΛ࣮ࢪ͠ɺ༗ޮੑΛ֬ೝ
ݕࡧߦಈͷྨ • Motivation • ߪೖͷҙࢥ͕ଘࡏʢނোͷަϓϨθϯτʣ • ߪೖͷҙࢥ͕ͳ͍ʢͻ·ͭͿ͠ʹγϣοϐϯάαΠτΛճ༡ʣ • Target Specify
• Known product category • Known product name/type • Known product brand • Undetermined • Search Strategy • Specific Keyword • EvaluateʢInspect a product’s specification, price, etcʣ • Compare • Browse(season sales, news arrivals, etc) • Direct Purchase(directly buy a target item) •
ɹݕࡧҙਤͷྨ • ϢʔβͷݕࡧҙਤΛ̏ͭʹྨ • Target Finding (TF) • ԿΛߪೖ͔ܾ·͍ͬͯͯμΠϨΫτʹཉ͍͠Λݕࡧ •
Λ࣮ࡍʹௐͯظ͍ͯ͠Δ͔Ͳ͏͔ΛνΣοΫ • ر௨ΓͷͳΒߪೖ • Decision Marketing (DM) • ཉ͍͠ͷδϟϯϧܾ·͍ͬͯΔ͕ԿΛߪೖ͢Δ͔ະܾఆ • ෳͷΛൺֱͯ͠ͲΕΛߪೖ͢Δ͔Λܾఆ • Exploration(EP) • ԿΛߪೖ͢Δ͔ܾ·͍ͬͯͳ͍ɺߪೖҙࢥ͕ͳ͍߹ଘࡏ • ͪ࣌ؒʹηʔϧͳͲΛݕࡧ • ੜϓϨθϯτΛݕࡧ • ϢʔβͷΞϯέʔτͷ݁Ռɺ97.6%(288/295) ͷϢʔβ্͕هͷ3ͭͷྨʹ͍ͣΕ͔ʹྨ͞ΕΔ͜ͱΛ ֬ೝ
ݕࡧͷҙਤ͕Ͳ͏ߦಈʹө ͞ΕΔ͔ • γϣοϐϯάαΠτͷݕࡧϩάΛجʹߟ • ओʹݕࡧ݁Ռͷϖʔδ( Search Result Page)ɺͷϖʔδͰͷ Event
log Λར༻ • Query string • Page number • Result list (set of product ids) • Clicked product ids • Add to cart event • Buy event
ηογϣϯϨϕϧͰͷੳ • ηϯγϣϯ͋ͨΓͷΫΤϦ • EP ͕ TF, DM ʹൺͯଟ͍ •
DM ͕ TF ʹൺͯᷮʹଟ͍͕େ͖ͳࠩͳ͍ • ݕࡧͷਂ͞ʢݕࡧ݁Ռͷϖʔδʣ • DM(4.88), EP(5.11) ͕ TF(2.91) ʹൺͯਂ͍ • Add to Cart, Buy Events • Buy Events TF, DM ͕ EP ʹൺͯଟ͍ • Add to Cart Event EP, DM, TF ͷॱʹߴ͍(ଈߪೖ͍ͨ͠ҙࢥͱൺྫ)
EventϨϕϧͰͷੳ • Click Event ͷճ • TF ͕࠷গͳ͘ɺηογϣϯ͋ͨΓͷ Click
EP ͕࠷ଟ͘ɺΫΤϦ͋ͨΓͷ Click DM ͕࠷ଟ͍ • ৄࡉͳใΛΓ͍ͨҙਤ͕͋Δ͔Ͳ͏͔ʹؔ • Click Event ͕ൃੜ͢ΔҐஔͷൺֱ • TF ্͕ҐͷҐஔͰͷ CTR ͕ߴ͍ •
Query String Ͱͷੳ • Query ͷ͞ɺTF ͕࠷͍ • TF ɺϒϥϯυ໊ϓϩμΫτͷಛͳͲΛμΠϨΫτʹΫΤϦʹؚΊΔ
͕ߴ͍ • ࠷ॳͷΫΤϦ͕࠷͍ • ηογϣϯதͷQuery ؒͷྨࣅੑ • Jaccard index Ͱൺֱ • EP ͕࠷ Query ͷྨࣅੑ͕͘ɺTF ͕࠷ߴ͍ • ෯͘ݕࡧ͢Δ΄Ͳ͘ͳΔ
User Behaviour, Search Intent and Satisfaction • ࣮ݧσʔλͷऩू • Experimental
Procedure • 6TF, 6DM λεΫΛ༩͑ɺQuery ͳͲ ఏࣔ • ຬͷ͕ݟ͔ͭΒͳ͍߹ඃݧ ऀ͕ Query Λมߋ • ݕࡧͷ݁Ռ࠷తʹ߹ͬͨΛબ ͼɺݕࡧͷຬΛ 5 ஈ֊ධՁ • ͦͷޙ 10 ؒϑϦʔͷݕࡧΛ࣮ࢪ(EP) ͠ɺຬΛ 5 ஈ֊ධՁ • Participants • େֶؔऀͷஉঁ֤ 10 ਓ
ຬͱߪങߦಈͷؔ • ຬ͕ߴ͍΄Ͳߪങߦಈ͕ݟΒΕΔ • ΧʔτʹೖΕͯߪೖ͢ΔϢʔβͷํ͕Χʔτʹ ೖΕͣʹߪೖ͢ΔΑΓएׯຬ͕ߴ͍ • ޙͷߪೖҙࢥ͋Γຬ͕ߴ͘ͳ͍ͬͯ ΔͷͰͳ͍͔ •
ຬ͕͍߹Ͱߪೖ͢Δέʔε͕ଘࡏ • ରͷʹͨͲΓண͘·Ͱʹۤ࿑͕͋ͬ ͨՄೳੑ • ߪೖ͠ͳͯ͘ຬ͕ߴ͍Ϣʔβଘࡏ • ߪങߦಈͷΈͰϢʔβͷຬΛਖ਼͘͠ ѲͰ͖ͳ͍Մೳੑ
ຬͱݕࡧૢ࡞ͷؔ • Query Level Statics • ຬ͕ߴ͍΄Ͳ࣍ͷ Query Λ͛Δ· Ͱͷ͕͍࣌ؒ
• ຬ͕ 5 ͷϢʔβ3,4 ͷϢʔβʹൺ Δͱਂ͍ݕࡧΛߦΘͳ͍ʹ͋Δ • తͷͷ͕͙͢ʹݟ͔ͭΔ • Event Level Statics • ຬ͕ߴ͍΄ͲΫϦοΫ͕ଟ͍ • ຬ͕ߴ͍΄ͲΫϦοΫͨ͠ͷϖʔ ʹཹ͢Δ͕͍࣌ؒ • ɹTF, DM ΑΓͦͷ͕ڧ͍
ຬͷ༧ଌϞσϧ • Feature • Query Level Feature: • ΫΤϦͷִؒ, •
ৄࡉݕࡧͷਂ͞ • Click Feature: • ΫϦοΫճ • ΫϦοΫҐஔ(avg, max, min) • landing page Ͱͷࡏ࣌ؒ(avg, sum) • ࡏ࣌ؒͱΫϦοΫͷΫϩε(Number of clicks with dwell time of 0-5/5-10/10-15/15-30/30-60 seconds) • Feature Λجʹɺclassifier Λ࡞Γ classifier ͷੑೳɺಛྔͷॏཁɺҙਤ(TF, DM, EP)ຖͷϞ σϧͷੑೳൺֱΛ࣮ࢪ
Prediction Results • Classifier ͷੑೳൺֱ • ϢʔβຖʹαϯϓϦϯάΛ͠ɺΫϩεόϦσʔγϣϯͰධՁ • GBDT ͷ݁Ռ͕࠷Α͍
• ಛྔͷॏཁ • GBDT ͷ Importance ͕ࢦඪ • ΫΤϦִ͕ؒҙਤʹΑΒͣ࠷ॏཁ͕ߴ͍ • ࡏ࣌ؒͷ߹ܭҙਤʹΑΒͣॏཁ͕ߴ͍ • ΫϦοΫΠϕϯτʹؔ͢Δಛྔॏཁ͕͍ • ҙਤຖͷϞσϧੑೳͷൺֱ • TF,DM EP ΑΓਫ਼͕ߴ͍ • EP ໌֬ͳߪങҙࢥ͕ͳ͍Ϣʔβ͍ΔͷͰɺͦͷΑ͏ͳϢʔβ ͷಛΛัΒ͑ΒΕ͍ͯͳ͍෦͕͋Γࠓޙͷ՝
·ͱΊ • ΦϯϥΠϯγϣοϐϯάʹ͓͚ΔϢʔβͷݕࡧҙਤΛఆٛ • TF ໌֬ͳҙਤΛ࣋ͬͯݕࡧΛߦ͍ɺߪങ·Ͱͷߦಈ͕গͳ͍͕ߴ͍ • DM ෳͷΛൺֱ͢ΔͨΊɺ͍ΫΤϦͰͷݕࡧΛߦ͍ɺͷৄࡉϖʔδΛݟΔ͕ߴ͍ •
EP ෳͷΫΤϦΛ͛ͯ৭ΜͳΛӾཡ͢Δ͕ৄࡉϖʔδ͋·Γݟͳ͍͕ߴ͍ • ࣮ݧʹΑΓϢʔβͷຬͷ༧ଌϞσϧΛ࡞ɺධՁ • TF, DM ͷਫ਼͕ಛʹߴ͍ • ຬͷܭଌΛϢʔβ͔ΒͷϑΟʔυόοΫͷΈͰܭଌ͓ͯ͠Γɺଞͷܭଌํ๏͕ඞཁͳ Մೳੑ͕͋Δ • ϢʔβʹΑͬͯຬɺෆຬͷײ͡ํ͕ҧ͏ • ͦͦຬɺෆຬΛ͖ͬΓ͍֮ࣗͯ͠ͳ͍߹͋ΔͷͰͳ͍͔