taus open source machine translation showcase, paris, sándor sojnóczky, hunnect, 4 june 2012

Post on 29-Nov-2014

716 Views

Category:

Technology

1 Downloads

Preview:

Click to see full reader

DESCRIPTION

This presentation is a part of the MosesCore project that encourages the development and usage of open source machine translation tools, notably the Moses statistical MT toolkit. MosesCore is supported by the European Commission Grant Number 288487 under the 7th Framework Programme. For the latest updates, follow us on Twitter - #MosesCore

TRANSCRIPT

TAUS OPEN SOURCE MACHINE TRANSLATION SHOWCASE

The ups and downs of implementing an MT environment15:45-16:00Monday 4 June

Sandor SojnoczkyHunnect

22

The Ups and Downs of Implementing an MT Environment

2

English Hungarian

33

The Ups and Downs of Implementing an MT Environment

Introduction

Hunnect Limited (since 2003)

LSP based in Hungary

Sándor Sojnóczky, MITI (linguist, economist, certified translator and university lecturer)

44

Background

Experience - 10 years, over 120 million words

Service requirements - Translation-Editing-Proofreading

Languages - Mainly Eastern European (35+)

Volumes – the number of words are constantly growing

Business Model – Faster–Better–More Economical

The Ups and Downs of Implementing an MT Environment

55

The Ups and Downs of Implementing an MT Environment

What changed in the past 10 years?

Service requirements – mostly TEP, but…

Languages – no significant changes in composition

Volumes – growing

Business Model – same direction, but…

5

66

The Ups and Downs of Implementing an MT Environment

The focus increases on…

Faster – Better – More Economical

Translation Automation

6

77

The Ups and Downs of Implementing an MT Environment

Challenge - Can we implement it as a business model?

Yes, if we consider 3 factors and calculate accordingly:

• Translation velocity• Pricing• Quality (TEP)

88

The Ups and Downs of Implementing an MT Environment

Translation(10 days)

Editing(3 days)

Proofing(2 days)

MT(1 day)

Post-editing(5 days)

Proofing(2 days)

Business Model Analysis – Timeframe/velocity

(based on a 25,000 word project)

Free (7 days)46% saving in time

C

o

s

t

Timeframe

99

The Ups and Downs of Implementing an MT Environment

Business Model Analysis – Pricing

40-50%

45-60%

5%

30-35%

5-15%

25-35%

5%

20%

Margin

Proofing

Editing

Translation

Margin

Ling. QA

Post-editing incl. fuzzies

Machine trans.

1010

The Ups and Downs of Implementing an MT Environment

Business Model Analysis – Pricing (Rewarding the post-editor)

TR PE TR/PE%

2,500 5,000

1.00 0.60 - 40%

+ 20%3,0002,500

Daily output

Unit paid

Total paid

1111

The Ups and Downs of Implementing an MT Environment

Successes

• Selection of the MT environment and collection of the domain-specific training data

• Training strategy for would-be post-editors

• Implementing MT as a business model

12

The Ups and Downs of Implementing an MT Environment

Challenges

• Retraining the MT engine– Collecting feedback from post-editors

– Handling lower quality data for the retraining

• Engaging post-editors– Linguists vs. domain-specific experts– Cloud/crowd

• Business model– Pricing

– Velocity

• New languages

1313

The Ups and Downs of Implementing an MT Environment

Thank you for your attention!

For further information and questions please feel free to contact:

Mr. Sándor Sojnóczky (ssandor@hunnect.hu)

Web: www.hunnect.hu

Web: www.hunnectacademy.com

top related