Esteam Translation Technologies and Systems

1
Esteam Translation Technologies and Systems

• QualityFast translation system
• Apollo translation system
• FlowManager work-flow management system

2
The Translation Market

• The language translation industry will be
• over 15 Billion in 2002
• 250 to 350 million pages per year translated
• Growing at 25 to 30 annually
• Only 8 of all translatable material is actually
  translated
• 17.9 billion pages archived without translation

3
What is Machine Translation?

• Machine translation is a complex technology that
  translates text from one human language to
  another.
• Computers interpret the structure of sentences in
  the source language and generate a translation
  based on the rules of the target language.
• The process involves
• breaking down complex and varying sentence
  structures
• resolving ambiguities
• synthesizing the information into the components
  and structure of the new language.

4
What is Translation Memory?
Translation Memory - an accessible database
of previously translated sentences, words and
expressions Advantages - highly accurate
for similar materials if a data
resource exists. Disadvantages - less context
sensitive than human
translation
5
Machine Translation vs. Human Translation

• Human Translation
• Benefits
• Can make judgement call
• Specialize in topic areas
• Accuracy
• Disadvantages
• Speed about 2,000 words per day
• Cost - 40 per page and up
• Availability

• Machine Translation
• Benefits
• Speed - Up to 20,000 words per minute
• Cost - pennies per word
• Disadvantages
• Language Availability
• Accuracy
• Cant make human judgement calls

6
Machine Translation Applications (1)

• Publishing - Dissemination of Information
• Types of Documents
• Documentation
• Multi-lingual Web sites
• Localization
• Professional translators translate, edit and
  update documents for high quality results.
• MT tools provide 30 productivity gain to human
  translators

7
Machine Translation Applications (2)
Gisting - Assimilation of Information MT
solutions empower anyone with the ability to
translate information quickly and easily. Types
of Documents Web Pages Intranets email Disadvan
tages Low accuracy / confidence levels Can
misinform / mislead Lead to costly and
inappropriate actions
8
Translation Memory Applications

• Published documents
• Document and manual localization
• Web page origination
• Legal document translation

9
Machine Translation Process
10
Integrated Translation Memory and Machine
Translation

• Combines the power of both worlds
• Highly accurate
• Very customizable
• Can be easily maintained and improved

11
ESTEAM Translation Systems Advantages

• Traditional Systems Approach
• One size fits all
• Speed more important than quality
• No system control once translation has started
• Dictionary building as the only quality control
• ESTEAM Approach
• System parameters tailored to the application
• Quality more important than speed
• User has full control via tool kit
• Several flexible quality control features

12
ESTEAM QualityFast System
How do we solve the speed / quality trade-off?

• Stable translation platform
• Highly accurate, context sensitive
• dictionaries and translation memory
• Dynamic monitoring of translation quality
• Dynamic editing of dictionaries and
• translation memory

13
Stable Translation Platform

• Core technology proven over seven years.
• Used by prestigious clients in government,
  publishing, legal and security applications.
• Process of continuous upgrade of lexical
  components and technologies.

14
Dictionary Innovation

• Dictionary and Translation Memory Management
• System (DMS) linked to an Oracle database.
• Ability to store and retrieve words,
  expressions,
• sentences and sub-sentences.
• Dictionaries and Translation Memory can be
• dynamically customized for and by every client.
• Dictionaries can have up to any level of context
• sensitivity

15
Dynamic Monitoring of Translation Quality

• Each sentence presented for translation.
• One or more dictionaries may be used for the
  translation, and the sequence set by the
  user.
• Resultant translation displayed on screen,
  together with specific parameters on
  translation quality.
• The sequence and/ or use of dictionaries may be
  changed at any time by the user.

16
Dynamic Editing of Dictionaries

• Powerful and flexible structure.
• Ability to recognize words, expressions,
  sentences and sub-sentences.
• Ability to create translation dictionaries from
  mono-lingual text data.
• Ability to capture data produced by other machine
  translation and translation memory systems.
• Ability to enhance dictionary during the
  translation process.

17
QualityFast Translation Process
Segment Source Text
Formatted Source Text
Translate sentences / sub-sentences
Save Format
TM and DictionaryDatabase
Translation Selection Rules
Verify Translation
Language Statistics
Format Translated Text
Translated Text
Combine translated units
18
File Formats For Translation

• Optional Formats for Files
• HTML, SGML, XML
• Microsoft Word (RTF), WordPerfect (RTF),
  PageMaker (RTF)
• Interleaf (ASCII),

19
Available Language Directions (1) All Languages
Link to All
English
Any Language
German
To Any Other Language
French
Spanish
Portuguese
Swedish
Danish
Finnish
Norwegian
Italian
Dutch
20
Available Language Directions (2)
English
Swiss German
German
French
Brazilian Portuguese
Portuguese
Finnish
Swedish
Danish
Spanish
Dutch
Flemish
Mexican Spanish
Italian
Norwegian
21
Available Language Directions (3)

• More than 100 language directions today
• Dialect sensitive
• Domain sensitive
• Rapid creation of new languages / directions
• Unparalleled quality performance

22
ESTEAM QualityFast System

• Speed
• About 7,500 words/min for unformatted text
• About 5,000 words/min for formatted text
• 700 Mhz Pentium class machine.
• Quality
• ESTEAM offers state of the art quality
• high accuracy - depending on source data
• Improvements gained with domain dictionaries
• Platforms
• Windows 2000, NT
• UNIX / AIX / Sun O/S

23
ESTEAM QualityFast System Typical Application
Areas

• Applications where quality is paramount
• Translate Enable the entire Enterprise
• Provides local or remote client access to a
  server-based translation engine.
• Internet and Intranet Environments
• Includes the tools to enable translations of
  plain text and document files over a networking
  environment
• Productivity application
• Translation services for corporate and
  government users

24
ESTEAM QualityFast System Typical Application
Areas

• Case Study
• CDT Unit of European Community
• Professional Translators
• Very high volume of translation
• 110 language directions
• Target of 80 saving on translation costs

25
QualityFast Translation Process(1) Internal
Translation Management
26
QualityFast Translation Process(2) Translation
Memory below sentence
27
QualityFast Translation Process(3) Concordancer,
Lexicon and TM Management
28
QualityFast Translation Process(4) Lexicon
Search Management
29
QualityFast Translation Process(5) Translation
Memory Search Management
30
QualityFast Translation Process(6) TM Search
Management - Sub-Sentence
31
ESTEAM Apollo Translation System

• Typical application targets
• Where fast, reliable message delivery over-rides
• top-quality translation output.
• Web Page translation for internal usage
• E-mail message translation
• International, intra-corporate, memos.
• Emergency and security notices
• Short dwell-time information announcements

32
ESTEAM Apollo Translation System
Speed About 12,500 words/min for unformatted
text About 8,000 words/min for formatted text IBM
RX6000 machine. Quality ESTEAM offers good
quality output High Accuracy - depending on
source Improvements gained with domain
dictionaries Platforms Windows NT UNIX / AIX /
Sun O/S
33
ESTEAM Apollo Translation System
Render clarify, depict, interpret, rephrase,
represent, translate. Rendering translation
with speed to give a clear and accurate
interpretation of the original Accuracy 75 -
90 Gisting translation of key elements where
speed is more important than accuracy. Accurac
y 50 - 80
34
ESTEAM Apollo Translation System

• Case Study
• CompuMark - Division of Thomson Corp.
• Translation of legal texts
• Very high volume of translation
• Multiple language directions
• No human intervention
• 24/7 operation

35
ESTEAM FlowManager System

• Functional Components
• The FlowManager product consists
• of several functional components.
• FlowManager Engine
• Translation Queue Manager (TQM).
• Translation Service Manager
• The Client Component

36
ESTEAM FlowManager System

• Translation Queue Manager
• The Translation Queue Manager provides an
  interface between the outside world and the
  Translation Service.
• It acts much like a printer queue, accepting
  requests from a variety of sources and monitoring
  the status of the Translation Service to ensure
  that the requests get acted upon and the
  requestor is notified of completion.

37
ESTEAM Translation Systems Characteristics

• Runs on Windows Servers
• Standard protocol access - Sockets/HTTP
• Scalable performance
• Clients for text, web pages and files
• Browser and standalone versions of clients
• Interface is documented
• Clients are easily customizable

38
Client Customization Process
Lexicon Manager
Translation Management Process
Selection Rules and Statistics
Translation Memory
Translation outputs
39
Translation Process Scalability
FlowManager
Translation inputs
QualityFast
Translators
External Trans.
Revisors
Translation outputs