translation technology and workflowmanagement · clauses, participle clauses, ellipses,...

1

Terminology and TechnicalDocumentation

Translation technology and workflowmanagement

Frieda Steurs - 2007 - TSS

Structure of the presentation

Source text controlIssues in translation and multilingualdocument management :

Case 1 : Integrated Language Services (ILS)

Case 2 : Cummins Engine CompanyCase 3 : Tweddle Litho CompanyCase 4 : Yamagata EuropeCase 5 : Atlas Copco

Conclusion

2


Some aspects of source text creation

Management and usability of technicaldocumentationThe reader/user :

UnexperiencedExpert

‘Restructuring minimalism’The minimal manual


Source text management

Case 1 : documentation for air linereservation system

Research into user behaviour :80% reduction in volume of the manual80% reduction of phonecalls to helpdesk

Case 2 : documentation for parceldelivery

74 billion $ losses due to the use of an inadequate Operational Manual

3


Controlled language

Constrained terminology, syntax and/or semanticsClear and consistent styleEnhance clarity, usability, transferability, retrievability, extractability, and translatabilitySome resistance when first using itThe objectives are met


How to Providethe right information

at the right time

in the right form

relevantaccuratenon-ambiguous

retrievableup-to-datemaintainable

clearstructuredconsistenttranslatable

4


Content Optimization Measures

Controlled Language and CL CheckerSpelling and Grammar CheckerText-graphics ratioSingle-sourcingMeta-informationTranslation MemoryMachine TranslationProcess Control...


What is Controlled Language?

Controlled Language is a subset of conventional language.

The subset is characterized bya standardized terminology a restricted core vocabularya restricted set of grammar and style rules

5


Motivation

Readability and international understandability: one element in a series of text optimization measuresTranslatability: bridge the gap of a deficient AI statusNational and international standardization in the context of concurrent writing and single-sourcingRetrievability in the context of knowledge and content management systems


Controlled English : History

1935 Ogden: BASIC-English1965 Caterpillar CFE1981-1986: AECMA Simplified English1986-1999:

• Rank Xerox• Eastman Kodak• IBM• General Motors• Lucent Technologies• …

Plain English MovementOther Languages such as French and German

6


ReadabilityCore VocabularySyntax

GovernmentAdministrativeDocuments

PlainEnglish

ReadabilityConsistencyTranslatability

Terminology(CoreVocabulary)Syntax

OtherIndustries

Complex Manufacturing

ICT

TechnicalDocuments

ControlledEnglish

ReadabilityConsistency

(Terminology)Core VocabularySyntax

AeronauticIndustry

TechnicalDocuments

SimplifiedEnglish

PurposeStandardizationLevel

SettingApplicationLinguisticVariety

Classification Scheme

Standardized Terminology

7


What is Standardized Terminology?

Principle One: 1 Concept = 1 Term⇒ disallow synonyms, orthographic and

morphological variants


Principle One :1 Concept = 1 Term

Synonyms

GSM

handy

mobile

wireless phone

Morphologicalvariants

indexes

indices

GSM

indexes

8





Principle Two: 1 Term = 1 Concept⇒ disallow homonyms


Principle Two : 1 Term = 1 Concept

Homonyms

airbag

noise

application

airbagairbag system

unwanted soundelectronic interference

piece of softwareresult, action of applying

9






Principle Three: 1 Term = 1 Part of Speech⇒ disallow homographs


Principle Three : 1 Term = 1 Part of Speech

Homographstest

abrasive

absent

nounverb

nounadjective

adjectiveverb

10



Principle One: 1 Concept = 1 Term⇒ disallow synonyms, orthographic and morphological variants


Principle Three: 1 Term = 1 Part of Speech⇒ disallow homographs

Principle Four: Acceptance of Term⇒ prefer terms that are linguistically adequate, transparent in

meaning and common in use


Why StandardizeTerminology?

TERMS = BASIC KNOWLEDGE ELEMENTSTerms help you to organize, distribute and standardize

knowledgeComprehensibility: common understanding of documentsTranslatability: only one correct translationMaintainability: reduced set of terms and translationsConsistency: always the same terms for the sameconceptsRetrievability: unambiguous set of basic knowledge elements

11

Restricted Core Vocabulary


What is a restricted core vocabulary?

DO NOT USEaccomplish

utilize

endeavor

accelerate

encounter

detail

normally

USEdo

use

try

speed up

find

instruction

usually, correctly

12


Why Restrict the Core Vocabulary?

Negative evidence for international understandability

Reduces acceptability by writers more than standardized terminology or restricted syntax

In avoiding lexical restrictions important arguments against CL are dropped or reduced:

• Learnability problem

• Restriction of expressive power

Restricted Syntax

13


Prohibited FeaturesLong and complex sentencesMany noun phrases per sentenceMany nouns or adjectives in a rowMany different sentence structuresComplex verb forms and tensesCompressive structures such as attributeclauses, participle clauses, ellipses, contractionsInconsistent or wrong punctuation


ReadabilityNon-controlled:

Tighten the screw several turns, remove the screwdriver and check to see that the cover still is in place on the phone before connecting the phone wire.

Controlled:1. Tighten the screw several turns.2. Remove the screw driver3. Check to see that the cover still is in place on the

phone4. Connect the phone wire

14


Comprehensibility

Non-controlled:When fasteners are removed, always reinstall them at the same location from which they were removed.

Controlled:Always reinstall fasteners in the same location.


Translatability

Non-controlled:A 10 mega-ohm input impedance digital voltmeter

Controlled:A digital voltmeter with 10 mega-ohm input impedance

15


Non-ambiguity

Non-controlled:Unscrew the plug-in unit from the wall and paint it.

Controlled:Unscrew the plug-in unit from the wall and paint the plug-in unit.


Case 1 : ILSIntegrated Language Services

Printing businessLanguage related issues :

CopywritingTranslationTerminology managementTranslation memoryRemodelling and finetuning of workflows

16


ILS : Main Tasks

Coordination and streamlining of thetranslation processCreation, hosting and updating of translation memories and termbases (in-houseterminology)Editing, copywriting and localisation issuesContent creation for multilingual websitesSource text control and management


ILS : WorkflowPre-study of existing documentation

Style briefing of translatorsReference material evaluation by the customerTerminology study of source texts and mapping ofinconsistencies

Implementation of MAHTInternal and external evaluationFull project executionFollow-up and evaluation

17


Intercultural ProblemsAsian customers and the European marketFiles are in unknown or incompatible file formatSource text has been translated into‘Japenglese’Content is not adapted to European marketRewritten sourcetexts need local marketguidelines for translatorsText expansion


Knowledge management

Data/ Information / KnowledgeInformation :

TextualGraphicalCorporate Identity

Input : source controlOutput : a variety of formats

CD-ROM, Website, media-neutral databases, printed matter , etc.

18


WorkflowTD & TM Pretranslated

document

Translation & revision

Cleaning andupdatingTD/TM

Translateddocument

TD TM

Source text


Case 2 : Cummins Engine Company

Leading manufacturer of gas and diesel engines (4500 distributors in 130 countries)50 service manuals and 60 partsmanuals every year

19


Towards more efficient localisationprocedures

The publications had to be localised in a more efficient wayReduce publication time and translation costTools for :

Selection and output of source/target data fortranslationUse of translation memoryPost-editing by translatorsControlled language (authoring process)

More accurate matching in the translation memory


Source text managementSource language EnglishInformation chunks in manageable fragmentsDiffing algorithm : comparison of the last and current versions of the document todetermine the editing changesComparison in TM : unmatched items are sent for translation.Upgrade of TM with the new information in the translation and in the source document

20


Results

6 languages : Sp, Fr, G, I, Sw, PortOperator and Maintenance Manual (400 pages) : from 6-8 months to 10 weeks.65-70% reduction in translation costs :

Use and reuse of data across publicationsIntegration with terminology and TMBatch composition system for layout and graphics (saves 30% on production costs)


Case 3 : Tweddle Litho Company

Technical publishing house forautomotive industryComplete data management services including authoring, translation and other related services

21


Better, faster, cheaper

Ford Motor Company : 30 languages, 60 countries (localisation!) supporting the vehiclerelease.Lay-out , graphics, had to be more accurateCulturally neutral global format and meet local requirements for engineering, regulatory, safety and environmentalconditions. Reduce publication cycle time and costs


The solutionAuthoring process was changed : SGML fragments in a document management systemText and graphics were separated and storedin dedicated databasesReduce page countMore accurate dataImprove production cycleOne process applies to all marketsSupports late product changes

22


Key to the solution

Data model and DTD : SGMLInformation granularity / storageMedia specific output : paper / electronicEffectivity control : target audience, type of vehicle involvedDocuments and CDs in over 30 languages. Increase in shared data up to 90%. Turnaround time for all markets from 6 months to 2 weeks.


Problems with the information flow

The use of flat files without data reuse acrosspublicationsInconsistent information from one publicationto anotherPublications had to become portable to a variety of software applications and output tomultiple media formatsTranslators had to start each manual fromscratch, and this made the production cyclevery long

23


Document management system

SGML for datamarkupInformation portability acrossapplicationsInformation was chunked into smallfragments :

Data reuse across publicationsEasy updates to all publications using these fragments


Content management

Content management systems are systems that capture, archive, index, manage, author, combine, link, and distribute internal and external information to create a knowledge repository

24


The need for a centralised process

Write it onceReuse it in many placesTranslate it onceReuse it in many places


Yamagata Europe

Automated Quality Control in TechnicalTranslation

25


Yamagata EuropeTranslation work : 100% outsourced30.000.000 words translated in 200530 different language pairs13 different formats700 translatorsSeveral specialisations

Vendor management: selection and evaluation : important for a language service providor


ControlReliabilitySize of the translatorsgroupTransparancyLack of good and specialised translatorsCapacityConfidentiality

Problems (1)

26


Problems(2)

Lack of timeSize of the translatorsgroupNo inhouse-expertise for every language pairManual revision is both financially and practically not feasible


Problems (3)

Subjectivity of the existing evaluationcriteria(e.g. SAE J2450, LISA QA model,…)

Based on manual revisionArbitrary and subjective catagoriesNo consensus

27


Translation Quality ControlNeed for quality control

Need for automated processes:

Formal mistakes in translation work are most of the time indicative for other problems in the text

“Everything that can be measured, can be traced”Distinction between formal mistakes and stilistic

(subjective) mistakes


QA Distiller

Error categories:OmissionsInconsistenciesFormatting issuesTerminology mistakes

28


Vendorselection Online registration

Candidate

Test translation

Approved

First project

On Trial

Assigned to account


29



30



31



32


Vendor testing

A. Processrelated translation qualityCommunicationPunctualitySegmentationTagging


Vendor testing

B. Productrelated translation qualityOmissionsInconsistentiesFormattingTerminologySpelling

33


Case 5 : Atlas Copco

Atlas Copco : industrial and professional toolscompressed air equipment, construction and mining equipment etc.(26.000 employees worldwide)

Bowne Global Solutions :Management of the technicaldocumentation and localisation of thecompressor product information


Atlas Copco : the challenge

Specialised manuals in more than12 languagesConditional publishingSource once, Write many times

34


Atlas Copco manualsPublish instruction manuals on the Web in web standard XML format - requiring no specific software from userPersonalisation of the instruction manual in 3 ways:

q conditional publishing to the specific equipment (no more/no less info): all information maps in master documents in repository are provided withretrieval tags, corresponding to model and its features

q in the language of choiceq with metric or British units of measurement

Manual content is defined by sales dept and customerA few mouse clicks automatically convert retrieved XML file to PDF and e-mail it to user


Atlas Copco manuals

Each manual is dynamically generated on-the-fly, allowing instantaneous update of contentAllowing decentralised and fragmentedtechnical authoring and translationReduce/eliminate time-to-market, pre-press, paper, printing, distribution cost

35


Schematical representationS ystem S ystem arch itectu rearch itec tu re

Im port trans la tions

S e lectto be

transla ted

S u bm it/C heck-in .

TriD K

R eposito ry

M apsM aster docum ents

IllustrationsSchem a,tem plates

P ub lishpe rsona lized

docum ent

to W eb

to pd f

In te rface withC R M / p rod con f.

sys tem

W ebW ebclien tc lien t

X M L X M L trans l. trans l.

C rea te lin k

C heck-ou t /E xpo rt

X M L X M L auth .auth .


Translations

12 EC languages Project allows increase to 18 languagesTrados translation software:q builds up translation memories per language pair for utmost re-use

of already translated sentencesq enables Multiterm terminology management, forcing external

translators to use pre-determined terminology, improving quality

q allows pre-calculation of translation cost (once translation memories are built up)

36


ConclusionOptimal procedures in technical document creation and translation :

Source text controlTerminology management (both source text andtranslation oriented)Translation managementContent managementCritical analysis of the needs of the userWorkflowmanagement