sam joachim

Translation memory systems to Translation memory systems to enhance the quality and enhance the quality and

productivity of localization of productivity of localization of teaching materialsteaching materials

Sam Joachim

6th Workshop Software Engineering Education and Reverse Engineering, Ravda (Nessebar), Bulgaria, 18 – 23

September 2006

Last years suggestion for future Last years suggestion for future development of S-Bahn Tooldevelopment of S-Bahn Tool Transformation of our PowerPoint material in an

independent XML format (perhaps based on <ML>3) ◦ With support for different learning paths ◦ With the possibility of easily exchanging parts of the content

(case studies, examples, …) ◦ With different outputs (ppt, pdf, html, textbook …)◦ For different target groups, perhaps with diverse knowledge levels

Building of a kind of repository / pool system for the learning objects and an authoring system for adaptable study packs or (e-)learning material

Support the use of ‘Authoring by Aggregation’ in the system

2Translation Memory Systems

Authoring by aggregation: Authoring by aggregation: main ideamain idea

A new learning object is composed of some parts from LO A (a picture), LO B (some text) and 2 new modules Na and Nb


Learning object A

Learning object B

New learning object

Information

Objects / Modules

Modul Na

Modul Nb

Some problems to solveSome problems to solveLearning object models adapted to

‚Authoring by aggregation‘

XML representation of our slides / material

S-Bahn tool problems with respect to localization (quality, efficiency of translation)


Learning object models Learning object models adapted to ‚Authoring by adapted to ‚Authoring by aggregation’aggregation’

&&

XML representation XML representation of our slides / materialof our slides / material


o

n

n

n

e

ts

w

y

seee

yt

ss

w

g

Learnativity Content Model Learnativity Content Model (Duval & Hodgins 2003)(Duval & Hodgins 2003)Sh

apes

, Pict

ures

, Tex

tfiel

ds, D

iagr

ams

77Translation Memory SystemsSource: E. Duval and W. Hodgins, A LOM research agenda. In Proceedings of the twelfth international conference on World Wide Web, pages 1–9. ACM Press, 2003.

Modular Content Hierarchy

JCSE CourseSection, Topic

Slides, Section

Grouped elements, single or

associated slides

Modular content objects Modular content objects explained on a typical slideexplained on a typical slide


Information object / Remark for the audience

Learning Object / Group of elements

Raw media object / textfield

Raw media object / textfieldRaw media object

/ textfield

Generation of fine-granular Generation of fine-granular objectsobjects

Use / adaptation of an existing XML teaching material language

Automatic generation of◦ Raw media objects (shapes, textfields)◦ Information objects (groups of objects, graphics)

Semi-automatic by selecting from the automatically generated elements◦ Learning objects (associated slides, ..)◦ Higher level objects like

Aggregate assemblies (topics) Collections (whole JCSE)


Example for generated Example for generated objects objects


Waterfall model

Teaching / Learning Object or Material Repository

Schematic of a transformation Schematic of a transformation process:process:

knowledgegrid

.doc

.pdf

.ppt

eLesson

.

.

.

Essence

What is the essence?Text? Pictures? Style?

Tool Tool 2

Automated.Some AI?

Half-bakedessence

Tool 3 Final document

Interactivity.Some NI

XML?

Moodle

Ahyco

Something

Somethingelse

.

.

.

Tool 4

Automatic generation of “Raw media

objects“

Raw data and media elements in XML format

Information objects (groups of objects, graphics)

Automatic grouping of

connected objects

Objects in some Teaching Material Language

(LMML /<ML3>)

Interactiv „Authoring by Aggregation“ process

uses fragments/modules to generate new material

Two ways for representing ppt Two ways for representing ppt in xml in xml to preserve our ppt to preserve our ppt layoutlayout


embedding original ppt code in container elements in the xml-language<learning object>

<container ppt>

</container ppt>

</learning object>

binary ppt representation

<learning object>

<slide Heading=‚The classical waterfall model (1970)‘>

<graphic name=‚Waterfall Model‘>

<textfield pos=(10,40) … >

Analysis and Definition

</textfield>

</graphic>

</slide>

</learning object>

build a transformation from the elements from ppt to the elements in the xml-language and vice versa (easier modification of extisting material)

Translation Memory Translation Memory SystemsSystems


15

◦ Word by word translation◦ Repeated translation of

phrases◦ Difficult use of the

feature for the content slides / slides with reappearing content

S-Bahn Tool problemsS-Bahn Tool problems

Possible solution:◦ The use of a translation memory system.

Translation Memory Systems

TM-Systeme: IntroductionTM-Systeme: Introductiontwo main types of translation

support

◦Machine Translation (fully automatic) Try to translate autonomous, in most

cases nonsatisfying

◦Machine Assisted Human Translation (computer-aided) Mainly with Translation Memory Systems


Background - HistoryBackground - HistoryWho: Target group - professional translatorsWhere: professional translation agencies /

specialized companiesWhat: Big translation projects with lots of different

media and/or documents and dicument versions◦ Documents to translate:

Technical / project / product documentation Texts with technical /natural science

background Software in different language variants and

during different program versions Including GUI elements, user / product documentation


TM SystemTM SystemIs a data base

Records sentences or word groups (segments) with the corresponding translation

During translation, the TM-System searches the already translated segments for similarities with the actual segment

The translator can easily use already translated segments


TM System (cont.)TM System (cont.)Every sentence / segment will be checked

during translating of other texts

If it is already in the data base, it is possible to adopt the translation as it stands◦ No segment /sentence should be translatet two

times. (Similar to the S-Bahn Tool content slides feature, but much more generic in practice.)

Highly effective TM Systems ease routine jobs of the translator. He/she can concentrate more on the creative task of translating.


Fuzzy searchesFuzzy searches In realworld texts, a sentence or seqment is very seldom

exactly repeated. The most TM systems search not only for exact matches

but also for matches with a certain similarity, the so called Fuzzy Matches

Fuzzy Matches: only marginal differences, e.g. numbers , names, additional words…

Of course, the translator has to do a manual adaptation when taking over a fuzzy match

Example from JCSE Topic 3 Slide 26:Already known: Part of the phase ‚Analysis and Design‘, in which the basic use cases of the systems will be detected: use case diagramsNew: Part of the phase ‚Analysis and Design‘, in which the basic classes of the problem will be detected: class diagrams


Existing TM SystemsExisting TM Systems Trados

◦ http://www.trados.com SDLX

◦ http://www.sdlint.com Transit

◦ http://www.star-ag.ch/products/transit DéjàVu

◦ http://www.atril.com OmegaT

◦ http://www.omegat.org/ Transolution

◦ http://transolution.python-hosting.com/


http://www.trados.com/

http://www.trados.com/

http://www.sdlint.com/

http://www.sdlint.com/

http://www.star-ag.ch/products/transit






http://www.atril.com/

http://www.atril.com/

http://www.omegat.org/



http://transolution.python-hosting.com/



Parts of the TM SystemParts of the TM System


Translation Memory DB Terminolog

y / Glossary DB

Alignment Tool

Inserts already translated texts

GUI

Dictionary functions and sometimes glossary

Collects and provide technical terms during translation

Can easily changed and expanded

Quick online search

• with sentences / segments and their translations

Alignment Tool: Example Alignment Tool: Example GPLGPL


Typical uTypical user interfaceser interface

Two windows for source and target language

Fuzzy index with matchesTerminology dictionary


TM Sy

stem:

TM Sy

stem:

Transo

lution

Transo

lution

26

Text to translate

Suggested translation (from

fuzzy match) to be adapted by the

translatorFuzzy match: explanation of the

suggested translation Display of the

context of the text to translate


27

TM System: TM System: OmegaTOmegaT

Text to translate embedded in its

contextSuggested

translation (from fuzzy match) and

Glossary


TM System: TransitTM System: Transit

28

Text to translate

Suggested translation (from fuzzy

match) to be adapted by

the translator

Fuzzy match: explanation of the

suggested translation

Terminology


Terminology/Glossary

Seach:

Functions

Translation:

Fuzzy match:(45%) Specification of the structure of the software (50%) specification of components and their relationsSpezifikation der Struktur der Softwa Spezifikation der Komponenten und ihrer Beziehungen Terminology: Specification of components; relation

Other languagesMacedonian: Специфицирање на структурата на софтверот (софтверска архитектура), специфицирање на компонентите и нивните односиRomanian: Specificarea structurii SW (arhitectura software ), specificarea componentelor şi a relaţiilor între eleSerbian Cyrillic: Спецификација структуре софтвера (софтверска архитектура), спецификација компоненти и нјихових веза

Possible User Possible User InterfaceInterface

Use of a TM system for Use of a TM system for localization of teaching localization of teaching materialsmaterialsApplicabe especially for PPT materials because

◦ High number of reappearing text segments ◦ Not too much complete sentences, more phrases, segments

Higher support for translators higher productivityNot restricted to PPT, open for any other document

format Possibility for a TM system web service for Software

Engineering materials

Nessesary adaption: not only two languages but many


Thank You.Thank You.

Space for QuestionsSpace for Questions