Glossary & Resources

Glossary

Basic terms for localization

• language - a system of communication used by a particular country or community. ISO 639 is the main standard used to define language codes.
• locale - the implementation of a language in a given market, including formatting (numbers, date, etc), common expressions and cultural differences. For example, French France (fr-FR) is different than French Canada (fr-CA). BCP 47 is the main standard used to define locale codes.
• Internationalization (i18n): a set of best practices and design process that ensures that an application can be adapted to various locales without requiring code changes.
• localization / l10n - converting a program to run in a different locale. Most of the effort revolves around translating UI string text, so "localization" often gets used synonymously with "translation". But technically also includes designing different layouts (ex: scripts that prefer top-to-bottom right-to-left) and UI widgets (some icons flipping for right-to-left languages)
• CAT (Computer Assisted Translation) tool - an editor that is designed for translators to be efficient to use and integrated with other l10n services. The CAT tool UI usually has a 2-column interface in which each message's source (original) and target (translation) text are kept vertically aligned with each other.
• TMS (Translation Management System) - a workflow system that manages the end-to-end work of translation. Includes user upload and download, cost estimation & billing, distributing work to translators, integration of reviewers and secondary reviewers, QA / issue management, and post-editing. Some TMSes provide their own integrated CAT tools. In other cases translators choose to use their own CAT tool, in which case they may use an industry standard format like XLIFF to download the translation source and upload their finished translation.
• post-editing - after the user receives their translated document as the result of the main translation workflow in the TMS, the user may want to make their own final touches to the translated doc. Those final touches are called post-editing.
• Translation Memory (TM) - a database of previous translations (translation entry = source string, source language, target language, target string). TMs typically store individual messages as source strings in separate entries. TMs can be shared globally, shared within a company, and/or private to a single user.
• Machine Translation (MT) - letting an automatic translation program perform translation of the source text. This is usually performed only when no entries in the Translation Memory exit that match the source text. The reason is that it is usually easier & cheaper to start translation by correcting Machine Translation output than to write out the translated string from scratch.
• XLIFF (XML Localization Interchange File Format) - a localization industry standard file format that defines the structure for translation task data.
• Okapi Framework - a software framework that enables people to develop their own l10n software (CAT tools / TMSes). Hierarchy of classes is similar to XLIFF data hierarchy spec. Supports XLIFF and many common textual document file formats via a plugin-style architecture. Most CAT tools / TMSes are built on Okapi.
• translatable unit / text unit - the first level of granularity in which a translation document is broken down to in XLIFF/Okapi. Represents translatable text -- text to be translated by a translator. Usually corresponds to paragraphs, but depends on the file format handler implementation
• segment - a sub-unit of a text unit. Usually corresponds to a sentence. Also represents translatable text.
• source locale - language of the source text (input) for a translation.
• target locale - language of the translated text (output) of a translation.
• placeholder - a piece of information inline within a segment that should not be translated. Usually represented in the UI as an indivisible widget (or substring).
• placeholder type - which linguistic category that a placeholder represents, as it pertains to the work of translation. Common ones are 'gender' (of a word or person) and 'plural' (a number).
• Natural Language Generation (NLG) - a process that transforms structured data into natural language.

Terms for message formatting

• API - any function call(s) invoked by the user to perform message formatting.
• API argument format - syntax of values passed to input of message formatting API. May also refer to structure of values represented by the syntax. May be similar or same as authoring format. See message syntax.
• application locale - the locale for formatting (or formatting resources) requested by an application.
• authoring - writing a message by hand that adheres to some syntax.
• authoring format - syntax of message formatting inputs when constructed for a program, either manually (by a developer) or programtically (ex: a WYSIWYG tool used by a translator). May also refer to structure of values represented by the syntax. May be similar or same as API argument format. See message syntax.
• data model - a syntax-independent description of the structure of values passed to the message formatting API.
• implementation - code written to make the API achieve the intended behavior of output for a given input.
• interchange format - syntax/file format used to convert the inputs of message formatting into inputs for other systems (ex: l10n systems).
• interpolate - inserting the contents of one string in the middle of another, at places indicated by a pattern / placeholder. See translation merging.
• locale fallback - offering a reasonable substitute locale when the requested locale's resources are not available. Results may vary depending on context (ex: audio vs. text vs. video).
• locale matching - computing the locale fallback.
• resource - files bundled with an application that are loaded in by the executable code. UI strings, etc. and their locale-specific translations are typically stored as resources.
• translation merging - in l10n TMSes, the document-level interpolation of translated content. In other words, the replacing of translatable units in the source document with their equivalent translated units. See interpolate.
• selector - see placeholder type.
• specification - the rules we decide that describe what is passed to the API for message formatting (structure of data, syntax, etc.).
• serialization - how to convert in-memory representations of data to/from a file/stream.
• syntax - a general term to describe a set of rules that describe the set of allowed symbols and their ordering. Can apply to data, source code, communication protocols, etc. regardless of interface (stream, file). Sometimes used synonymously for file format.
• variant - one of the pre-defined values (cases) in a switch/case manner that a message can take depending on the value of some variable (switch), such as a polaceholder.

Synonyms

• AST (Abstract Syntax Tree) - the tree structure created by a parser of an input stream/file according to a particular syntax/format, typically in refernece to source code (as opposed to data files).
• binding syntax - see authoring format.
• build/parse-time format - see authoring format.
• compound message - a message that can take on different pre-defined values (cases) in a switch/case manner depending on the value of some variable (switch), such as a placeholder. See variant.
• consumed format - see API arugment format.
• developer format - see authoring format.
• DOM overlay - a way to enable merging translated HTML attributes for an HTML tag that is inline with the translated text. See interpolate and translation merging.
• file format - a standard syntax, coupled with a semantics for interpretation, to describe the contents of a file. Most commonly used for files representing data (including documents) and executables. See syntax.
• filter - Okapi terminology for the serialization code for a file format between input documents and Okapi in-memory data structures.
• formatting locale - the locale provided by the locale fallback mechanism for formatting a message.
• fragment message - when a message is used to represent a portion of a larger message, often nested within the larger message. Fragment messages can occur in messages with multiple variants as a means to refactor and narrow the region covered by variant text. See variant.
• full message - when a message is not nested with other messages (ex: the message is just an interposing of strings and placeholders). See fragment message.
• language negotation - see locale fallback.
• locale chain - see locale matching.
• intermediate format - see authoring format.
• markup - a category of file formats for plain text data in which portions of text are annotated with metadata / attributes. Boundaries of annotated portions are marked inline using text 'tags' that that are distinguishable from the main text. See syntax and file format.
• message syntax - the syntax of the inputs to message formatting. If the structure significantly changes between authoring and the API calls at runtime, and the representation used for notation also must differ, we can split this into API argument format and authoring format. The term syntax here may also refer to structure of values represented by the syntax.
• multi-level filter - Okapi terminology for a filter that supports the proper extraction of text units when contents adhering to one file format are embedding within contents adhering to another file format.
• placeable - see placeholder.
• placeholder locale - if placeholders contain content that is computed based on the locale, and if that locale is allowed to differ from the rest of the text in the message, then this is the locale for just that placeholder.
• positional variable - when a function call takes a series of values without some way to name those values (ex: comma-separated list). Changing the order of the inputs to the function call would lead to different semantics. In contrast, maps and named paramters are ways to provide arguments to function calls that are order-independent.
• selector - see placeholder type.
• source code representation - see authoring format
• standard message format - see message syntax.
• translation/localization format - the interchange format used specifically for l10n use cases. See interchange format.
• resource locale - see target locale.
• runtime format - see API argument format.
• UI language - see target locale.
• variable - see placeholder.
• variable locale - see placeholder locale.

Resources

• Localization Essentials (Udacity) - free