Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

machine translation manuel herranz PangeaMT TAUS Barcelona

12,883 views

Published on

how machine translation is about empowering users and how users can be empowered using DIY SMT technology to build their own statistical machine translation solutions

Published in: Technology, Business
  • Be the first to comment

machine translation manuel herranz PangeaMT TAUS Barcelona

  1. 1. - User empowerment -DIY SMT<br />Don't be afraid to provide the tools to those who need them<br />Manuel Herranz – PangeaMT - Pangeanic<br />www.pangea.com.mt<br />
  2. 2. Userempowerment<br />USERS<br />80% like<br />19% notlike<br />1% done before<br />http://t.co/HDTboxQ<br />
  3. 3. Userempowerment<br />USERS<br />80% like<br />19% notlike<br />1% done before<br />http://t.co/HDTboxQ<br />Meaning of USER becomingcloselyrelatedto COMMUNITY, <br /> POWER,<br /> FEEDBACK,<br /> ACCOUNTABILITY<br />
  4. 4. Humankindconstantsearchfor<br />TOOLS<br />better<br />more<br />otherthings<br />http://t.co/HDTboxQ<br />
  5. 5. Humankindconstantsearchfor<br />TOOLS<br />better<br />more<br />otherthings<br />http://t.co/HDTboxQ<br />Aninstrumentformaking material changesonotherobjects […]. Tools are theprimarymeansbyhumanbeings control and manipulatetheirphysicalenvironment – EncyclopediaBritannica.<br />
  6. 6. History of theworldlargelya fightto control<br />resources<br />tools[technology]<br />MT: Anothertranslatorout of business...... ?<br />
  7. 7. History of theworldlargelya fightto control<br />resources<br />tools[technology]<br />In 20th-21stcenturyalso a fightto control and manipulate<br />INFORMATION [data]<br />ACCESS [data]<br />
  8. 8. History of theworldlargelya fightto control<br />21st century<br />INFORMATION [data]<br />ACCESS [data]<br />IS THE ERA OF<br /><ul><li> SHARING
  9. 9. OPEN</li></li></ul><li>History of theworldlargelya fightto control<br />21st century<br />INFORMATION [data]<br />ACCESS [data]<br />* Communities<br />* Source (Linux, others)<br />* Data<br />IS THE ERA OF<br /><ul><li> SHARING
  10. 10. OPEN</li></li></ul><li>History of theworldlargelya fightto control<br />21st century<br />INFORMATION [data]<br />ACCESS [data]<br />USERS<br />havethepower<br />* Communities<br />* Source (Linux, others)<br />* Data<br />IS THE ERA OF<br /><ul><li> SHARING
  11. 11. OPEN</li></ul>“Wecannotsolvetheproblemusingthesametools and theway of thinkingthatcreatedit” A. Einstein<br />
  12. 12. MT at Pangeanic, from Trial toProduction<br />2007 and before<br />2007/08<br /><ul><li> RB tests withcommercial software
  13. 13. Insufficientlygood output
  14. 14. Onlyinternalproduction
  15. 15. EU Post-Editing Award
  16. 16. V1: Small data sets (2-5M words), </li></ul>automotive & electronics<br /><ul><li> (ES), thenFr/It/De in otherfields</li></ul>2009/10<br />.<br /><ul><li>Division born
  17. 17. 00's of enginetrials and languagecombinations
  18. 18. Open-Source to commercial
  19. 19. TMX / XLIFF workflows</li></ul>2011/12<br /><ul><li>DIY SMT
  20. 20. Empower Users
  21. 21. Glossary
  22. 22. Automated re-training
  23. 23. Transfer architecture and know-how to users
  24. 24. Compatibility with commercial formats (ttx, sdlxliff, itd)</li></li></ul><li>12<br />MT at Pangeanic, from Trial toProduction<br /><ul><li>Usersprovide information to improve [theyare the source & target]
  25. 25. Potential MT userswanted to beanotherPangeanic= buildtheirown systems
  26. 26. Somecan, somecan’t
  27. 27. Otherwantturnkeydevelopments
  28. 28. OtherspreferSaaS
  29. 29. Most want to unwrap the blackboxbutwithoutwalking the road</li></li></ul><li>PangeaMT<br />2009<br />2010<br />Predictions<br />Tech. notthe realm of afew providers<br />2011<br />2012<br />2013<br />User empowerment<br />000's of customized MT systems<br />2014<br />2015<br />YEAR<br />2016<br />2016<br />2017<br />2018<br />
  30. 30. PangeaMT<br />2009<br />2010<br />Predictions<br />Tech. notthe realm of afew providers<br />2011<br />2012<br />2013<br />User empowerment<br />000's of customized MT systems<br />2014<br />2015<br />YEAR<br />2016<br />2016<br />2017<br />2018<br />
  31. 31. PangeaMT<br />2009<br />2010<br />2017<br />2018<br />
  32. 32. PangeaMT<br />2009<br />MT acceptance<br />2010<br />Predictions<br />Tech. notthe realm of afew providers<br />2011<br />Until 2011<br />2012<br /><ul><li> MT acceptance growth.
  33. 33. Translator engagement challenge
  34. 34. Need for data has been addressed – still more work to be done.
  35. 35. Users and practitioners now can build their own systems. </li></ul>2013<br />Userempowerment<br />000's of customized MT systems<br />2014<br />2015<br />YEAR<br />2016<br />2016<br />In 5 years... after 2016<br />2017<br /><ul><li> Combinations??
  36. 36. Supra-engines??
  37. 37. World-knowledge??</li></ul> …...suggestions....???<br />2018<br />
  38. 38. PangeaMT<br />2009<br />2010<br />Summary<br /><ul><li>USER EMPOWERMENT : give people the tools so they can grow their own solutions
  39. 39. PangeaMT provides infrastructure
  40. 40. Cloud Training: so users concentrate in production, not in technical bits & updates
  41. 41. Pressurefordataavailabilitycoming fromusers will benefit efforts for standardization</li></ul>2018<br />
  42. 42. 18<br />Thankyou !<br />MANY QUESTIONS PLEASE!!<br />mherranz@pangea.com.mt<br />

×