Services which can be provided by a LSP in addition to translation.

This refers to the foundational systems developed by Asia Online to do generic translations in each language pair. The Baseline data is useful to augment data contributed for a specific domain and fill out vocabulary and linguistic knowledge gaps. It includes a wider range of vocabulary as well as the more obscure terms. Baseline provides some linguistic foundations to domain data and generally adds robustness to the SMT engine.

Asia Online prides itself in thoroughly cleaning the data prior to use in developing their baseline systems. Baseline data is constantly being improved and extended to add ever expanding coverage of vocabulary in different contexts.

BLEU is a measurement developed by IBM for evaluating the quality of text which has been machine-translated from one natural language to another.

Quality is considered to be the correlation between a machine's output and that of a human: "the closer a machine translation is to a professional human translation, the better it is". More detail is provided on BLEU on the Translation Quality Metrics  page.

BS EN 15038:2006 is the first European standard to set out the requirements for the provision of quality services by translation service providers (TSPs). It charts the best practice processes involved in providing a translation service through commissioning, translation, review, project management and quality control, to delivery.

The purpose of this European standard is to establish and define the requirements for the provision of quality services by translation service providers.

BS EN 15038:2006 encompasses the core translation process and all other related aspects involved in providing the service, including quality assurance and traceability.

This standard offers both translation service providers and their clients a description and definition of the entire service. At the same time, it is designed to provide translation service providers with a set of procedures and requirements to meet market needs.

Conformity assessment and certification based on this standard are envisaged.

BS EN 15038:2006 specifies the requirements for the translation service provider with regard to human and technical resources, quality and project management, the contractual framework, and service procedures.

The standard document can be downloaded from See http://www.bsigroup.com/en/Shop/Publication-Detail/?pid=000000000030122045 .

BS EN 15038 is published as the following national standards:

• PKN EN 15038 Poland
• BS EN 15038 Great Britain
• DIN EN 15038 German
• NF X50-670 France
• OENORM EN 15038 Austria
• SN EN 15038 Switzerland
• UNE-EN 15038 Spain
• SFS EN 15038 Finland
• UNI EN 15038 Italy

Data that has not been put through any cleaning process and is in its raw form as provided by the data supplier.

Some data is in a more "dirty" or "clean" state than other data. Generally all data should be scrubbed to a pristine state for best results.

Data that has been processed through Asia Online's tools to ensure that terms are normalized, formatting data, variables and other forms of data "dirt" that would reduce SMT system quality have been removed or corrected.

F-Measure is an algorithm developed by New York university for evaluating the quality of text which has been machine-translated from one natural language to another. F-Measure considers both precision and recall to compute the score. More detail is provided on F-Measure on the Translation Quality Metrics  page.

Rendering of spoken information in the source language into the target language in spoken form.

Language Model data tells the translation platform how to use words and phrases in the target language. It can be used to stylize and refine grammar and plays a key role in selection of the optimal output phrases.

An organization or company that provides translation and localization services for language related tasks.

The linguistic, cultural, technical and geographical conventions of the target audience

Localization is the transformation of information to the standards of other countries.

This includes the translation of all linguistic components (words, abbreviations, symbols, and units) into the target language. Considering software localization (an Added Value Service) this comprises translation of the user interface, online help, any displayed messages, strings in graphics, and documentation.

Additionally the adaptation of the content to the cultural standards of the country and the target audience is also performed for complete localization. Language use depends on context and the people involved. The way to communicate messages differs across countries and many cultural issues can arrise where a phrase or word in one country has a very different meaning in another country.

Localization should be carried out by or at least in part by professionals with a deep understanding of the target market and culture, not just the target language. To speak the language is usually not sufficient to localize a product, document or campaign.

Phrase Pair data is either full sentences or partial sentences that makes up a phrase. They are typically equivalent text in two or more languages.

Normalization is the process of organizing data and terminology in such a way that the results of translations are always unambiguous and as intended, while removing inconsistencies.

A runtime translation glossary (RTG) provides information to a translation engine at the time of translation that directly influences the current translation task.

Normally statistical patterns learnt from the training data are the key drivers to translation quality without a particular bias towards the documents specific terminology or exceptions.

Using an RTG allows for greater control of style, terminology and grammar, while also handling document specific exceptions.

Checking of proofs before publication.

Examine a target text for its suitability for the agreed purpose and respect for the conventions of the domain to which its belongs and recommend corrective measures.

Person who reviews.

Examine a translation for its suitability for the agreed purpose, compare the source and target texts, and recommend corrective measures.

Person who revises.

The original text. The text which is being translated.

The translated text. Text that has been translated from the Source Text.

Rendering of written text in the source language into the target language.

Because translation is a change process, and each change event introduces the possibility of further error, checks and balances are needed to ensure quality. An industry norm is a three-step process called Translate Edit Proof (TEP).

The TEP process starts by translating (Translate) the Source Text by a bilingual specialist or machine translation; it is then reviewed (Edit) for accuracy by a second translator who is familiar with the characteristics of your target audience and it is finally proofread (Proof) by the original translator or a third translator who approves eventual changes made and checks for flow.

Person who translates.

A translation memory (TM) is a database that stores segments that have been previously translated. The translation memory stores the source text and its corresponding translation in language pairs called “translation units”. A TM can store text in many forms including words, phrases and paragraphs.

See Language Service Provider (LSP) .

Tuning Set data is a set of “gold-standard” translation examples that the translation platform uses to determine the most optimal settings for high-quality translations during the system building phase.