©2012 tauyou <language technoloy> #1
Machine Translation for LSPs:
a practical guide
Diego Bartolomé, CEO
©2012 tauyou <language technoloy> #2
some pains
of human translation
clients pay less for higher value
difficult with fast turnaround times
quality cannot always be guaranteed
... and ... machine translation
is/was expensive
takes time to learn
quality / usability
©2012 tauyou <language technoloy> #3
where to start?
©2012 tauyou <language technoloy> #4
step 1: set realistic objectives (I)
why do we want machine translation?
increase revenue
reduce costs
open new markets
our clients are using it
because everybody is using it
just to play for a while
...
©2012 tauyou <language technoloy> #5
why do we want machine translation?
increase revenue
reduce costs
open new markets
our clients are using it
because everybody is using it
just to play for a while
...
step 1: set realistic objectives (I)
©2012 tauyou <language technoloy> #6
step 1: set realistic objectives (II)
a) increase revenue by 5% in 1 year by offering
new solutions to our existing clients
b) reduce costs in 10% by increasing the
productivity of our freelance translators
c) market a low cost medium-quality service in
Brazil for the language pair EN-BR with MT +
post-editing
d) become the leader in post-editing training
services for freelance translators
©2012 tauyou <language technoloy> #7
what are yours?
©2012 tauyou <language technoloy> #8
step 2: decide the requirements (I)
all possible language pairs we work with
in all the language domains of the company
in real time
with a 95% quality
service 24x7x365
paying less than 0.01% of our revenue
©2012 tauyou <language technoloy> #9
step 2: decide the requirements (I)
all possible language pairs we work with
in all the language domains of the company
in real time
with a 95% quality
service 24x7x365
paying less than 0.01% of our revenue
©2012 tauyou <language technoloy> #10
step 2: decide the requirements (II)
our top 3 languages to start with
possibility to extend it to the top 15
translation and do-not-translate glossaries
medical, legal, IT , and automotive domains
web service to handle all CAT tools formats
Word Error Rate below 30%
12x5 support
©2012 tauyou <language technoloy> #11
step 2: some hints to start with
your top languages
the most similar languages
as a stand-alone tool to handle any file format
the cleanest translation assets
the biggest translation memories
the innovative freelance translators
the clients willing to test/try
©2012 tauyou <language technoloy> #12
step 3: evaluation of solutions
open source – in-house or service company
Moses, Joshua, cdec, Marie, Apertium,...
online services
Bing, Google, ...
RbMT
Gramtrans, Lucy Software, Systran, ...
SMT
Asia Online, SAIC, SDL, tauyou, ...
©2012 tauyou <language technoloy> #13
step 3: evaluation of solutions
open source – in-house or service company
Moses, Joshua, cdec, Marie, Apertium,...
online services
Bing, Google, ...
RbMT
Gramtrans, Lucy Software, Systran, ...
SMT
Asia Online, SAIC, SDL, tauyou, ...
©2012 tauyou <language technoloy> #14
step 3: some hints
own solution or 3rd party
in-house or hosted
language combinations
is confidentiality an issue?
customization capabilities
improvements over time
evaluation reports
©2012 tauyou <language technoloy> #15
step 4: carry out pilots
testing is the only way to know if it works
gather all your translation assets
choose 3 alternatives and pay for a test
how it starts
how it improves
how it achieves the goals
the biggest cost is your time
©2012 tauyou <language technoloy> #16
step 4: some hints
test all posible alternatives in providers
try to measure results over a period of time
define your key metrics and track them
the improvement ratio is a key indicator
customization capabilities are relevant
advanced NLP capabilities
ask other clients about it
©2012 tauyou <language technoloy> #17
now, move on
©2012 tauyou <language technoloy> #18
step 5: ongoing processes
measure, measure, measure
translation time
translation costs
your preferred metric, e.g. Word Error Rate
client complaints
create new engines in new language pairs
©2012 tauyou <language technoloy> #19
a number of failures
unlimited testing offering leads to failed systems
not enough good data
difficult language pairs
technological issues
unclear expectations
...
detect it fast and pivot
failure makes you learn and improve
©2012 tauyou <language technoloy> #20
several successes as well
tourism site EN/FR
patent translation EN/FR
bicycle e-commerce site EN/DE
legal EN/PT,ES,FR,IT,CA,GA
automotive ES/RO
information Technology EN/ES, EN/KR, EN/HE
medical EN/ES
generic web service public institution
...
©2012 tauyou <language technoloy> #21
Questions?
©2012 tauyou <language technoloy> #22
more info at
http://speakerdeck.com/u/tauyoucom
©2012 tauyou <language technoloy> #23
Thanks!
// Diego Bartolomé, PhD
<address> C/ Les Planes 39 – 08201 Sabadell – Spain
<phone> +34 93 711 29 96
<cell> +34 670 331 225
<email> dbc@tauyou.com
<www> tauyou.com

More Related Content

PPTX
6 weeks Summer Training with live project on Dot Net, Java, PHP in Gurgaon D...
PDF
Odoo Implementation Methodology
PPTX
HgTechSolutions
PDF
Odoo Experience 2018 - The Odoo Methodology
PDF
Converted file 8a4a782a
PDF
CRM - Get Quality Leads and Opportunities
PDF
PDF
TAUS Post-editing webinar. Spanish-to-English Module
6 weeks Summer Training with live project on Dot Net, Java, PHP in Gurgaon D...
Odoo Implementation Methodology
HgTechSolutions
Odoo Experience 2018 - The Odoo Methodology
Converted file 8a4a782a
CRM - Get Quality Leads and Opportunities
TAUS Post-editing webinar. Spanish-to-English Module

Similar to 2012 GALA Webinar: Machine Translation for LSPs. A practical guide (20)

PDF
2011 TAUS Executive Forum Barcelona: Language Technology for optimum localiza...
PPTX
Cloud-based Translation Management Systems
PDF
Workshop on the tauyou machine translation platform
PDF
2013 ALC Boston: Your Trained Moses SMT System doesn't work. What can you do?
PPTX
Algorithms for the content industry
PPTX
Localisation World Dublin 2014 - Disruptive Innovation: Opportunities and Cha...
PDF
Machine Translation Master Class at the EUATC Conference by Diego Bartolome
PDF
2012 MosesCore GALA Monaco: Friendly Machine Translation
PPTX
Elia ND Munich 2013 - The Future LSP: Partnering for Innovation
PDF
Human quality machine translation
PDF
Seeing the Wood for the Trees in MT Evaluation: an LSP success story from RWS
PPTX
Translation in the 21st Century Webinar
PDF
2011 Tekom Wiesbaden: Implementation of a machine translation engine at CPSL
PPTX
Strata - Final_IB_02_17
PPTX
High quality machine translation short id-2706
PPT
Boselli & Silva Mixing Cat And Mt
PPT
Corp preso it 04012014
PDF
TAUS MT SHOWCASE, Microsoft Translator, Chris Wendt, Microsoft, 10 October 2013
PPT
Mixing Computer-Assisted Translation and Machine Translation
PPTX
Easyling at atc London
2011 TAUS Executive Forum Barcelona: Language Technology for optimum localiza...
Cloud-based Translation Management Systems
Workshop on the tauyou machine translation platform
2013 ALC Boston: Your Trained Moses SMT System doesn't work. What can you do?
Algorithms for the content industry
Localisation World Dublin 2014 - Disruptive Innovation: Opportunities and Cha...
Machine Translation Master Class at the EUATC Conference by Diego Bartolome
2012 MosesCore GALA Monaco: Friendly Machine Translation
Elia ND Munich 2013 - The Future LSP: Partnering for Innovation
Human quality machine translation
Seeing the Wood for the Trees in MT Evaluation: an LSP success story from RWS
Translation in the 21st Century Webinar
2011 Tekom Wiesbaden: Implementation of a machine translation engine at CPSL
Strata - Final_IB_02_17
High quality machine translation short id-2706
Boselli & Silva Mixing Cat And Mt
Corp preso it 04012014
TAUS MT SHOWCASE, Microsoft Translator, Chris Wendt, Microsoft, 10 October 2013
Mixing Computer-Assisted Translation and Machine Translation
Easyling at atc London
Ad

More from tauyou (20)

PDF
Artificial Intelligence and Machine Learning found in Translation
PDF
I can't help falling in love with machine translation
PDF
Mind the gap between what you say and what you deliver
PDF
Some Lessons Learned on Machine Translation
PDF
From the Lab to the Market
PDF
APIfying the Translation Industry
PDF
The Discreet Charm of Machine Translation
PDF
Women in Localization UK Webinar with Diego Bartolome
PDF
The Beauty of Machine Translation
PDF
Emerging Technologies Enabling New Business Models
PDF
Innovating in Translation
PDF
Pushing Machine Translation Forward
PDF
The State of Post-Editing
PDF
lo que he aprendido (y quiero compartir)
PDF
What you need to put Machine Translation into practice: Tools, People, and Pr...
PDF
How we failed to win a 100,000,000 word contract (GALA Istanbul 2014)
PDF
Learn to Innovate (GALA Istanbul 2014)
PDF
Entrepreneurship in Education
PDF
2013 UAB Barcelona: Change the world (one start-up at a time)
PDF
2013 Tekom Wiesbaden: A Business Model Generation Session
Artificial Intelligence and Machine Learning found in Translation
I can't help falling in love with machine translation
Mind the gap between what you say and what you deliver
Some Lessons Learned on Machine Translation
From the Lab to the Market
APIfying the Translation Industry
The Discreet Charm of Machine Translation
Women in Localization UK Webinar with Diego Bartolome
The Beauty of Machine Translation
Emerging Technologies Enabling New Business Models
Innovating in Translation
Pushing Machine Translation Forward
The State of Post-Editing
lo que he aprendido (y quiero compartir)
What you need to put Machine Translation into practice: Tools, People, and Pr...
How we failed to win a 100,000,000 word contract (GALA Istanbul 2014)
Learn to Innovate (GALA Istanbul 2014)
Entrepreneurship in Education
2013 UAB Barcelona: Change the world (one start-up at a time)
2013 Tekom Wiesbaden: A Business Model Generation Session
Ad

Recently uploaded (20)

PDF
Lung cancer patients survival prediction using outlier detection and optimize...
PDF
Early detection and classification of bone marrow changes in lumbar vertebrae...
PDF
The-2025-Engineering-Revolution-AI-Quality-and-DevOps-Convergence.pdf
PPTX
Internet of Everything -Basic concepts details
PDF
SaaS reusability assessment using machine learning techniques
PDF
LMS bot: enhanced learning management systems for improved student learning e...
PDF
A symptom-driven medical diagnosis support model based on machine learning te...
PDF
A hybrid framework for wild animal classification using fine-tuned DenseNet12...
PDF
Introduction to MCP and A2A Protocols: Enabling Agent Communication
PPTX
Build automations faster and more reliably with UiPath ScreenPlay
DOCX
Basics of Cloud Computing - Cloud Ecosystem
PDF
The AI Revolution in Customer Service - 2025
PDF
Ensemble model-based arrhythmia classification with local interpretable model...
PDF
ment.tech-Siri Delay Opens AI Startup Opportunity in 2025.pdf
PDF
Auditboard EB SOX Playbook 2023 edition.
PDF
Dell Pro Micro: Speed customer interactions, patient processing, and learning...
PPTX
Presentation - Principles of Instructional Design.pptx
PDF
Transform-Your-Streaming-Platform-with-AI-Driven-Quality-Engineering.pdf
PDF
IT-ITes Industry bjjbnkmkhkhknbmhkhmjhjkhj
PDF
Transform-Quality-Engineering-with-AI-A-60-Day-Blueprint-for-Digital-Success.pdf
Lung cancer patients survival prediction using outlier detection and optimize...
Early detection and classification of bone marrow changes in lumbar vertebrae...
The-2025-Engineering-Revolution-AI-Quality-and-DevOps-Convergence.pdf
Internet of Everything -Basic concepts details
SaaS reusability assessment using machine learning techniques
LMS bot: enhanced learning management systems for improved student learning e...
A symptom-driven medical diagnosis support model based on machine learning te...
A hybrid framework for wild animal classification using fine-tuned DenseNet12...
Introduction to MCP and A2A Protocols: Enabling Agent Communication
Build automations faster and more reliably with UiPath ScreenPlay
Basics of Cloud Computing - Cloud Ecosystem
The AI Revolution in Customer Service - 2025
Ensemble model-based arrhythmia classification with local interpretable model...
ment.tech-Siri Delay Opens AI Startup Opportunity in 2025.pdf
Auditboard EB SOX Playbook 2023 edition.
Dell Pro Micro: Speed customer interactions, patient processing, and learning...
Presentation - Principles of Instructional Design.pptx
Transform-Your-Streaming-Platform-with-AI-Driven-Quality-Engineering.pdf
IT-ITes Industry bjjbnkmkhkhknbmhkhmjhjkhj
Transform-Quality-Engineering-with-AI-A-60-Day-Blueprint-for-Digital-Success.pdf

2012 GALA Webinar: Machine Translation for LSPs. A practical guide

  • 1. ©2012 tauyou <language technoloy> #1 Machine Translation for LSPs: a practical guide Diego Bartolomé, CEO
  • 2. ©2012 tauyou <language technoloy> #2 some pains of human translation clients pay less for higher value difficult with fast turnaround times quality cannot always be guaranteed ... and ... machine translation is/was expensive takes time to learn quality / usability
  • 3. ©2012 tauyou <language technoloy> #3 where to start?
  • 4. ©2012 tauyou <language technoloy> #4 step 1: set realistic objectives (I) why do we want machine translation? increase revenue reduce costs open new markets our clients are using it because everybody is using it just to play for a while ...
  • 5. ©2012 tauyou <language technoloy> #5 why do we want machine translation? increase revenue reduce costs open new markets our clients are using it because everybody is using it just to play for a while ... step 1: set realistic objectives (I)
  • 6. ©2012 tauyou <language technoloy> #6 step 1: set realistic objectives (II) a) increase revenue by 5% in 1 year by offering new solutions to our existing clients b) reduce costs in 10% by increasing the productivity of our freelance translators c) market a low cost medium-quality service in Brazil for the language pair EN-BR with MT + post-editing d) become the leader in post-editing training services for freelance translators
  • 7. ©2012 tauyou <language technoloy> #7 what are yours?
  • 8. ©2012 tauyou <language technoloy> #8 step 2: decide the requirements (I) all possible language pairs we work with in all the language domains of the company in real time with a 95% quality service 24x7x365 paying less than 0.01% of our revenue
  • 9. ©2012 tauyou <language technoloy> #9 step 2: decide the requirements (I) all possible language pairs we work with in all the language domains of the company in real time with a 95% quality service 24x7x365 paying less than 0.01% of our revenue
  • 10. ©2012 tauyou <language technoloy> #10 step 2: decide the requirements (II) our top 3 languages to start with possibility to extend it to the top 15 translation and do-not-translate glossaries medical, legal, IT , and automotive domains web service to handle all CAT tools formats Word Error Rate below 30% 12x5 support
  • 11. ©2012 tauyou <language technoloy> #11 step 2: some hints to start with your top languages the most similar languages as a stand-alone tool to handle any file format the cleanest translation assets the biggest translation memories the innovative freelance translators the clients willing to test/try
  • 12. ©2012 tauyou <language technoloy> #12 step 3: evaluation of solutions open source – in-house or service company Moses, Joshua, cdec, Marie, Apertium,... online services Bing, Google, ... RbMT Gramtrans, Lucy Software, Systran, ... SMT Asia Online, SAIC, SDL, tauyou, ...
  • 13. ©2012 tauyou <language technoloy> #13 step 3: evaluation of solutions open source – in-house or service company Moses, Joshua, cdec, Marie, Apertium,... online services Bing, Google, ... RbMT Gramtrans, Lucy Software, Systran, ... SMT Asia Online, SAIC, SDL, tauyou, ...
  • 14. ©2012 tauyou <language technoloy> #14 step 3: some hints own solution or 3rd party in-house or hosted language combinations is confidentiality an issue? customization capabilities improvements over time evaluation reports
  • 15. ©2012 tauyou <language technoloy> #15 step 4: carry out pilots testing is the only way to know if it works gather all your translation assets choose 3 alternatives and pay for a test how it starts how it improves how it achieves the goals the biggest cost is your time
  • 16. ©2012 tauyou <language technoloy> #16 step 4: some hints test all posible alternatives in providers try to measure results over a period of time define your key metrics and track them the improvement ratio is a key indicator customization capabilities are relevant advanced NLP capabilities ask other clients about it
  • 17. ©2012 tauyou <language technoloy> #17 now, move on
  • 18. ©2012 tauyou <language technoloy> #18 step 5: ongoing processes measure, measure, measure translation time translation costs your preferred metric, e.g. Word Error Rate client complaints create new engines in new language pairs
  • 19. ©2012 tauyou <language technoloy> #19 a number of failures unlimited testing offering leads to failed systems not enough good data difficult language pairs technological issues unclear expectations ... detect it fast and pivot failure makes you learn and improve
  • 20. ©2012 tauyou <language technoloy> #20 several successes as well tourism site EN/FR patent translation EN/FR bicycle e-commerce site EN/DE legal EN/PT,ES,FR,IT,CA,GA automotive ES/RO information Technology EN/ES, EN/KR, EN/HE medical EN/ES generic web service public institution ...
  • 21. ©2012 tauyou <language technoloy> #21 Questions?
  • 22. ©2012 tauyou <language technoloy> #22 more info at http://speakerdeck.com/u/tauyoucom
  • 23. ©2012 tauyou <language technoloy> #23 Thanks! // Diego Bartolomé, PhD <address> C/ Les Planes 39 – 08201 Sabadell – Spain <phone> +34 93 711 29 96 <cell> +34 670 331 225 <email> [email protected] <www> tauyou.com