Artificial Intelligence: Cutting Down Translation and Technical Documentation Costs

letecký průmysl l služby l podpora výroby

throughout the entire document, which is

a very important factor for the overall quality of the text.

If the software does not find a sentence

that is 100% identical, it is able to find

a sentence that is at least similar; this is

where fuzzy logic comes in. An example is

a long sentence, where one or two words

change, while the rest of the sentence remains the same. In this case, the translator

receives a so-called fuzzy match, where

he or she immediately sees the differences in both sentences highlighted in colour. The translator can then use the offered translation and manually translate the

changed words. The result is, once again,

improved efficiency and better consistency throughout the translation of the documentation, i.e. improved overall quality.

The goal is repeated use of the translator's

work using a sophisticated comparative

mechanism, which results in a 30–70%

reduction of time and costs in the translation of technical documentation. The

function for maintaining consistent terminology, which is especially important in

technical documentation, also improves

the overall translation quality.

We must also mention a new technology, which has been developing over the

past few years at the speed of light. By

now, everyone has heard of Machine

Translation (MT), or has at least tried some of the available free online translating

tools. These often produce comical translations and cannot be used as a reliable

source of translation if exact formulations

are important, which technical documentation, contracts or marketing materials

certainly require. Translators do not need

to fear for their jobs but the manner in

which they work is certainly gradually

changing. After two decades of using

translation memories and fuzzy logic, specialized systems for machine translations,

customized for a specific area and a particular client, are more and more popular.

We have been developing the statistical

system of machine translation called STAR

MT, in cooperation with university professors of computer linguistics, for several

years now. We prepare individual algorithms from dual and single language entities and client terminology for large and

long-term projects. The result is a very efficient solution, where suggestions from

the translation memory and fuzzy matches are supplemented with machine

translation suggestions, which the translator checks and modifies as needed. Our

clients, thus, see a great reduction of time

and costs, while the quality of the translation is maintained.

A project needs to be of a specific size

and continuity in order for MT technology

Digital continuity

• Synchronized

exchange of data

between R&D, spare

parts management

and technical

publication via

interfaces to PDM

/ERP systems

• Digital consistency reduces the amount of effort

involved in research and data collection, thereby

reducing the time to market, and also closes

information gaps when it comes to making

constructive changes

• Obsolescence management services for the life

cycle of parts and components

Collaboration with partners

• Information

structures such

as S1000D are

standard for

documentation projects

in the area of aerospace

and defense

• GRIPS supports all common S1000D versions.

If necessary, information can be semantically

recorded, maintained and automatically output

in selected S1000D subsets

• With the GRIPS Collaboration Portal solution,

S1000D data modules can be validated,

consolidated, administered and published for

various customers by multiple consortium partners

Information in standards

Simplified English?

• Depending on the

project types and

customers,

production

information

is required in various

documentation standards,

e.g. S1000D or DITA

• With semantics, you can make information fit for

complete flexibility: Via authentic single sourcing,

multi-channel publications and smart services

can be generated automatically

• The regulated

language of

"Simplified English"

is standard for many

military and aerospace

projects

• Simplified English is fully

supported by all STAR Group products and can be

seamlessly integrated into all editing, translation

and publication processes

to be effective; between 500 to 1 000 pages of new text per year. The text should

be simple, factual and instructional, such

as texts in user manuals and technical documentation, where the objective is an

accurate and direct translation, as opposed to a literary and creative translation.

The original text needs to be well-written,

using unified terminology, formatting, style, abbreviations etc. The machine translation produces a word-for-word direct

translation. If this technology is to save time and costs, we cannot expect for the

translator to regroup and significantly modify the structure of the sentences when

editing the machine translation. We also

cannot expect the same results in all languages. Romance and Germanic languages yield good results, as do some Slavic

languages. But you can forget about Hungarian or Estonian. Today's MT systems

are not able to grasp the complexity of

these languages. Using a machine translation should be one of the links in the entire process, from the efficient and highquality creation of a resource text,

through the use of a translation memory,

fuzzy matches and terminology, to the application of machine translation and its

constant improvements through the import of new texts into the MT algorithm.

Another new trend is the optimization of

creating technical documentation. There are

modern software tools on the market, which

help authors of technical documentation

to articulate information that needs to be

included in the technical manual most efficiently. Intelligent software offers similar

texts from previous documentation based

on a half-written sentence, which the author can use as inspiration. A notification

function for prohibited terminology, with

a synonym suggestion, is also available.

All the functionalities, including morphology analysis, is also available to authors of

documentation in Czech. The result is documentation written in a unified style,

using consistent terminology, where the

final translation into target languages will

be faster and cheaper since the suggested

texts already have their translations in the

translation memory.

Whether you need translation for your

products as an exporter or translation for

products imported from abroad, you can

save a lot of time and money by making

the entire localization process more efficient. You can purchase the CAT software

and the tool for optimization of documentation creation for your documentation

department in the Czech Republic, and

provide training for your internal translators and authors. If you work with a translation agency, choose one that uses CAT

software and will let you take advantage

of its financial benefits.

Contact: Ing. László Jankovics

laszlo.jankovics@star-group.net

tel no.:+420 608 320 100