Hosted by Henri Broekmate (Lionbridge). Panelists Nathalie Dougall (Booking.com), Matt Romaine (Gengo), Bernie Hsu (Alibaba).
For many firms, global crowdsourcing is an unparalleled way to reduce fixed costs and dramatically enhance operating efficiency and scalability. Enterprise crowdsourcing and big data seem to be a good match, particularly as data-related work can often be broken down into tasks or projects. Managing resources and workflows at the task or unit level is at the heart of the localization industry’s expertise. Many suppliers are finding that they can easily offer adjacent services such as search relevance, sentiment analysis, data tagging, user generated content curation, transcription, and data enrichment, in a crowd model – typically in a private crowd. This panel will discuss whether localization service providers and language technology innovators are offering relevant crowd solutions. Using technology as a lens, we will also discuss whether there is a role for increased crowdsourcing as the adoption of MT drives increased post-editing work. Lastly we will exchange ideas about where the crowdsourcing opportunity sits in the framework of the “gig” economy, and whether it will thrive and grow, or crumble under the push for legislation and labor reform.
2. TAUS Annual Conference 2016
The Sentinel, Portland, OR
The Internet is Eating the World, and
the Crowd has the Seat of Honor at the
Banquet.
3. Problem Statement – Focus on Information Crowdsourcing
The Uberization of everything
For many firms, global crowdsourcing is an unparalleled way to reduce fixed costs
and dramatically enhance operating efficiency and scalability. Enterprise
crowdsourcing and big data seem to be a good match, particularly as data-related
work can often be broken down into tasks or projects. Managing resources and
workflows at the task or unit level is at the heart of the localization industry’s
expertise. Many suppliers are finding that they can easily offer adjacent services
such as search relevance, sentiment analysis, data tagging, user generated content
curation, transcription, and data enrichment, in a crowd model – typically in a
private crowd.
This panel will discuss whether localization service providers and language
technology innovators are offering relevant crowd solutions. Using technology as a
lens, we will also discuss whether there is a role for increased crowdsourcing as the
adoption of MT drives increased post-editing work. Lastly we will exchange ideas
about where the crowdsourcing opportunity sits in the framework of the “gig”
economy, and whether it will thrive and grow, or crumble under the push for
legislation and labor reform.
6. Intro & Setup.
● Translations & Content Agency
○ Translations, Photo and Copy
● In-house language teams + freelance translations workforce (6+ years)
○ scaling out photo (own set of challenges!)
● Crowdsourcing photo and destination UGC
7. Deadline: May 2017
- 1 year preparation
- rolling out new agreements now
Impact on Freelancer:
● minimum weekly time requirements halved
● diversified income source
● regular confirmation
● more freelancers in ‘pool’ therefore more competition for regular work
Impact
8. Business impact:
● increased operational management
● existing freelancers to sign new agreement (also non-NL)
● recruitment, compliance, quality, productivity, payments
● + active regular follow up on workforce
● language teams - quality checking/training
Impact
9. We empower talented individuals with access
to new opportunities and reach across the
language barrier. We've raised $25MM from
international investors, and our “HQ” is on the
Internet. When our users go global — they
#gengoit.
Customers platform
37
languages
20,000+
Qualified Translators
Order translation from $0.05/word
Avg. 89minutes
Platform-oriented
From 1 word to millions per
week
sales@gengo.com | gengo.com
Translators
10. “Crowdsourcing” — why?
Customer
• Fire-and-forget
• Redundancy
• Rapid scaling
Translator
• Active — learning
from peers
• Passive — learning
from tech
• More control
11. Components of a good crowd platform
• bulk user-management
• group demotion; target group messaging
• activity monitoring — at various layers
• scalable infrastructure
• swarms — of work & workers
And readiness to iterate — quickly!
12. PRIVATE & CONFIDENTIAL - FOR INTERNAL USE ONLY
Alibaba Language Service
Allow billions of users to have better language Services
DIGITAL MEDIA & ENTERTAINMENT CORE COMMERCE LOCAL SERVICES
MARKETING SERVICES
DATA MANAGEMENT PLATFORM
CLOUD COMPUTING
LOGISTICS
PAYMENT &
FINANCIAL SERVICEData Collection
Data Application
Human Translation
Machine Translation
13. PRIVATE & CONFIDENTIAL - FOR INTERNAL USE ONLY
ALS Crowdsourcing
QUALITY SPEED
COST
Buyers and suppliers have different considerations on quality, cost and speed
14. PRIVATE & CONFIDENTIAL - FOR INTERNAL USE ONLY
ALS Crowdsourcing in core commerce
ALS
Human Translation
Or
Machine Translation
Post Editing
Marketplace
Alibaba/Aliexpress
Tmall Global
Release
Products
Suppliers
Buyers
Buyers
Buyers
One part of
Crowd-sourcing
15. PRIVATE & CONFIDENTIAL - FOR INTERNAL USE ONLY
ALS Crowdsourcing in local services
ALS
Human Translation
Or
Machine Translation
Travellers
Services Providers
Communication
Or
Shopping
Travellers Travellers
SPEED
Crowd-sourcing
16. PRIVATE & CONFIDENTIAL - FOR INTERNAL USE ONLY
ALS Crowdsourcing Overview
1、Higher requirements on speed;
2、Increase volume of post-editing work;
3、 Higher demand for crowdsourcing with more freelancers’ participation
Editor's Notes
Thanks Peng Wang, Lecturer and CAT Tool Coordinator from the University of Maryland, who has kindly consented to interpret for Bernie Hsu from Alibaba
Nathalie has been with Booking.com since 2010 and manages freelance operations in their Translations & Content Agency. Having joined the Freelance team at its inception in order to set up and organise operations, she has facilitated its growth from ‘working at home’ internal translators to a global network of freelancers working 24/7 in over 40 languages. With a view to scaling and operationalising further content products, photography and UGC are also part of her focus. Since recent changes in the Dutch law, Nathalie has been working hard with legal teams to re-write their freelance agreement, trying to ensure operations are not too disrupted! The implications of labor law changes for the freelance crowd and business operations are evident. The challenge is to keep things running smoothly with a flexible and productive workforce, keep the 'crowd' motivated and still find a balance between cost-effective delivery and scalable operations.
Matthew Romaine is founder and CEO of Gengo, a global people-powered translation platform empowering talented individuals to access new opportunities and reach across the language barrier. A technology enthusiast, Romaine is driven by a mission to help people connect and communicate easily around the world. Gengo was launched in 2008 with the goal of providing a simple tool for individuals to request human translations and soon after developed the first people-powered translation API, enabling businesses of any size to globalize and communicate freely. Prior to Gengo, Romaine was part of Sony’s R&D group, where he researched the future of audio and served as a key member in the Corporate Technology Department, developing growth strategies. Romaine holds a Bachelor of Arts in both Computer Science and Music from Brown University, and a Masters in Music, Science and Technology from Stanford University. He currently resides in Tokyo, Japan and is fluent in Japanese. Matt’s super passionate about his mission, a passion that he will communicate clearly in his introductory presentation
Bernie Hsu is Operations Leader of Alibaba Language Services (ALS). Starting in April of 2008, Bernie worked for Baidu Network Technology (Beijing) Co., Ltd., and Alibaba.com as a product manager. In the beginning of 2014, he took charge of product design and operations of the Alibaba Language Services.
Lux Arumque
Light and Gold
Light, light
Light, light
Light, light
Warm
Warm
And heavy
And heavy
And heavy
Pure
Pure
Pure like gold
They sing and sing and sing
To the newborn babe.
Anne-Maj, do you have an example of a best-in-class presentation used to frame a panel discussion?
20 minutes for introductions
15 minutes for panel discussion
10 minutes for audience interaction
Do we want to use slides?
Three or four minute intro by the host
Would depend on how many minutes? Matt
Bernie: agrees with Matt
Nathalie: potential buyer of the crowd – presentation about positioning and what we’re looking for?
New Yorker technology cartoons
Crowdsourcing behind all of the apps referenced, i.e. search relevance, identity management, trust relationships
Questions:
The concept of groups of people working on a single project is not new; why has crowdsourcing in the language industry become a discussion topic these past few years?
What can translation buyers do to get the most out of crowdsourced translation platforms?
Is our panel discussion limited to information crowdsourcing?
Matt: our presentations cover the question whether LSPs and language technology companies offer relevant crowd solutions;
Bernie: agrees with Matt…introduction about Alibaba answer will answer the question;
Nathalie: one of the things we’re trying to solve is crowdsourcing for translation moderation
Nathalie: MT will drive more crowdsourcing….makes sense to Matt, also Bernie application of crowdsourcing for MT PE, especially for mobile applications
Matt: I haven’t been following the regulatory framework….do you have some links?
Nathalie: legislation in The Netherlands has changed….Dutch government have abused using a freelance workforce…major impact on my operation
Bernie: different crowdsourcing environment in China
Is translation / L10N per se a form of information crowdsourcing?
Who wants to frame the core question, i.e. that LSPs, because of the nature of their work, are well positioned to expand into information crowdsourcing?
Does anyone have a good graphic to communicate the technology landscape in crowdsourcing?
Who wants to tackle the MT angle?
The question about the gig economy and the regulatory framework?
Using information crowdsourcing and machine learning – we’re focused on information crowdsourcing today
Translator
- active: forums; discussions
- passive: TM, resources, lurking
- more control: choose the work; choose the payment cycle; choose the availability (no “mindshare” issues / being forgotten by pjmers for hiatus); focus on work and not negotiating
- managing 20,000+ users means building user-agnostic / bulk feature-sets