Transcript Shopping Across the Language Barrier

Shopping Across the Language
Barrier
Workshop on Multilingual Data Value Chains
in the Digital Single Market
Brussels, 16 Jan 2015
Dave Lewis, ADAPT @TCD
[email protected]
ML eCommerce Trends
• 23 Million European SMEs, create 85% new jobs
– Finding customers single biggest problem
– International SME show more jobs growth (7% vs 1%) and
innovation (26% vs 8%)
• eCommerce growing globally:
– CAGR 18.3% to 2018, 2.2% of European GDP 2013
– EU: 50% shop online, only 20% cross border and growing only
at ¼ rate
• ML eCommerce: SMEs must
– Must excel in Niches: product specific terms and knowledge
– Engage with customer online: two-way conversations
Example:
Micro-SME ecommerce value chain
Customer Engagement Ecosystem
Content
Social
Analytics
Translation
Media
Search
Online
Trade Guilds/
&
SEO
Communities
Associations
Knowledge &
Events
training resources
Wholesalers
Wholesalers
Wholesalers
Wholesalers
Niche Value Add
Decoupage.ie
ecommerce
SaaS
•
ePayment
Service
Digital Single Market works well downstream
–
–
–
•
•
Customer
English as lingua franca
Export-focused Medium-SME wholesalers and service providers
MNCs with establish multilingual offering
Language Barrier firmly in place upstream
Challenge to Customer Engagement Ecosystem
–
–
Must become systematically multilingual
Must serve micro-domains that allow SMEs to add value
Challenges ML Data Value Chain
• NLP (MT, term extraction, WSD) can help
language barrier in customer engagement
• BUT MUST
– Scale across languages (& pairs)
– Support niche language used in micro-domains
• Technical terms “Washi Tape”, “Glitter Glue”, “Crackle
Varnish”
• Product Names “Daily Art”
• Community usage and neologisms “Over-decoupaged”
– Reflect community need for discourse, protect
competitive knowledge
– Respect copyright and data protection rules
Research Challenges
• Leverage professional knowledge
– Instrument/guide language worker to optimise NLP training CNGL, FALCON
• Build on massively multilingual lexical-conceptual
resources
– BabelNet, 13 Million entries, 270+ languages
• Linguistic Linked Data - LIDER, META-SHARE
– Common meta-data, collective intelligence, discovery, quality,
licensing
• Integrate Micro-Domain LT into Customer Engagement
Ecosystem – social media & ecommerce platforms
– Autocomplete (especially m-commerce)
– Autotag, ML SEO
– Micro-domain MT
• THANK YOU!