what you need to put machine translation into practice: tools, people, and processes
DESCRIPTION
You have no doubt been hearing about machine translation (MT) for several years now, but translating it into a high Return on Investment is not an easy task. We have heard many success stories presented at conferences, but we know a high number of LSPs that have failed to achieve their goals. In this master class we will share some success stories but also some failure use cases, so that we can all learn from them. After that, three hints on tools, people and processes will be given to help you successfully integrate MT into your workflow. Finally, we will discuss several examples of post-editing English machine translation output from various source languages.TRANSCRIPT
What you need to put Machine Translation
into practice: Tools, People, and Processes
Diego Bartolome @[email protected]
60+ clients
16 countries
~1 billion words in 2013 (cost reduction)
All language pairs
Why Machine Translation?
Why Machine Translation?
Price pressure
Tight deadlines
Client requests (impossible projects)
New needs from clients
Everyone else is doing it
...
Machine Translation requires
Initial investment
Time
Continuous improvement
Change in company culture
Clear goal
Machine Translation Companies
and others
What do you look for?
Some ideas
Confidentiality Customization
Control User friendliness
Flexibility Technology
Business model Quality
Relationship structure Trust
Engine building FAQ
Engine building FAQ
How many words/segments do I need?
How do I measure quality?
What language pairs work best?
What are the solutions if quality is unacceptable?
How often engines need retraining?
Process integration
What is the best process?
Most extended process
Prepare files
Pre-translate (>85%)
Prefill remaining segments with MT
Post-edit MT + review TM
Update TM (+MT) + compute quality metrics
Automatic Post-editing Rules
People
What are the challenges you have faced?
People
Averse to change
Importance of less time consuming errors
Training
Age/years of experience
Time vs. money
3 hints
Tools
invest in choosing the right one for you
Process
decide a process from the beginning
People
make everybody win
Post-editing examples
Change before you have to
Jack Welch
Other interesting topics
Optimized big data approach to MT
Quality Checks
Online postediting interface
Process automation (API)