Tools 3-99 (September 1999) by Sue Ellen Wright

So long as technical writers and translators are competent professionals to begin with, the single most significant controllable factor in controlling text and translation quality is consistency in using terms. Whenever multiple authors or translators are working on a large document or set of related documents, inconsistency in using certain terms may occur, even if each person in the document production chain is using some kind of authoring or translation tool that includes a termbase. The problem that can arise is that in today’s distributed environments, with freelancers and consultants working together on large projects, not everyone is necessarily using the same tools with the same termbase.

Additionally, more and more technical communicators and translators are moving into integrated environments involving multiple computer applications, particularly in the localisation area. Here we see an increasing need to leverage terminological resources across platforms and program boundaries.

The language industries are rapidly embracing the use of translation tools such as automatic terminology lookup, terminology mining, terminology consistency checkers, and machine translation. Authoring tools that involve access to a termbase are also appearing, at least in the context of controlled language, but will over time no doubt also be used in the authoring processes where the syntax is less controlled.

The danger that arises in these kinds of environments is that different tools use different kinds of terminology management systems. Certainly, everyone is familiar with integrated systems like - STAR Transit/TermStar and TRADOS Workbench/MultiTerm; products which enable users to access human-oriented terminology databases while using Translation Memory. But the criteria and modeling features of NLP dictionaries prepared for Machine Translation differ significantly from the concept-oriented terminology resources favored by human translators and multilingual technical writers.

The ideal terminology management system of the future will

• utilize "smart" extraction utilities to lay the groundwork for efficient term entry into databases, and
• be interoperable with a wide range of tools, not just with word-processing and TM, but with MT, various localisation tools, and AI expert systems, including ontologies.

A glance at existing interchange formats designed to make this all possible may cause confusion. A litany of acronyms confronts the potential user: there are MARTIF and GENETER, TMX and TBX, and OLIF, OTELO and OSCAR (which sounds like some sort of operatic trio where an errant Swede meets a noble, but tragic Moor and they end up serenading the heroine together with her perky, trousers-role page...). And they are implemented in SGML, XML, and stand-alone tagging formats that are similar to SGML. It’s alphabet soup with a strong dash of T’s, X’s, and O’s, not to mention M and L.

For a quick rundown, here’s who’s who:

• MARTIF, the Machine-Readable Terminology Interchange Format, is ISO 12200:1999; it is an SGML-based formalism intended for the interchange of terminological information between and among termbase systems.
• GENETER is a generic model for SGML-based representation of terminological data. Both MARTIF and GENETER are designed to
  1. represent terminological data and
  2. to serve as an intermediate format between different applications and platforms.
  They differ in that GENETER is designed for use with standard SGML parsing tools, whereas MARTIF is forward engineered with the intention of adapting it directly to XML schemas as soon as the XML committee finalizes the schema standard. Both standards have their adherents, with GENETER finding favor with some term bank systems and MARTIF (and its spin-off formats) looking attractive to language industry players who are interested in short DTDs and using schemas for data element and content validation.
• OLIF the (Open Lexicon Interchange Format) is a component of the OTELO project (Open Translation Environment for Localization) and was primarily intended for the exchange of lexical resources among different translation tools, especially in conjunction with OTELO's Common Lexical Database (CLDB). A major focus of OLIF is the expression of basic NLP-oriented terminological information used in MT environments. OLIF can accommodate very simple terminological entries, but it lacks MARTIF and GENETER’s ability to express embedded relationships and to mirror entry and termbase structures.
• A product of LISA, the Localisation Industry Standards Association’s OSCAR SIG (Open Standards for Container/Content Allowing Re-use), TMX, the Translation Memory Exchange Format, is rapidly being adapted by all major TM publishers to facilitate the interoperability of the various TM applications. OSCAR’s future TBX (Termbase Exchange Format) is likely to be a flavor of so-called "Blind MARTIF," which will be designed to enable people to simply use the format without customised negotiations with exchange partners. TMX is especially interesting because it was one of the first exchange standards written form XML and incorporates Unicode and meta-markup features that can be adopted to future interchange solutions.

All this makes for a nice stew of acronyms, but without an interoperable unifying core, these elements will not contribute adequately to system integration. Recently yet another group was formed called SALT: Standards-based Access service to multilingual Lexicons and Terminologies. SALT’s purpose is to

1. test and refine an XML-based lex/term-data interchange format combining MARTIF and OLIF and called XLT,
2. develop a website for people to try out various XLT utilities, and
3. develop an XLT toolkit for lex/term-related product developers.

The SALT consortium, under the leadership of Alan K. Melby (Brigham Young University (BYU) and LinguaTech International), has enlisted the cooperation of a range of American and European universities, as well as the interest of tools developers (Trados, Star, EP, Logos, Systran, and L&H), industry organisations (LISA, AMTA), and a number of LISA-affiliate corporations, such as Microsoft. It will continue it’s efforts under the aegis of the International Organisation for Standardization (ISO) Technical Committee 37 for Terminology, Subcommittee 3 for Computer Applications.

For more information on SALT and the SALT team, check out http://www.ttt.org/salt.

"transline Deutschland - Übersetzungsdienst für technische Übersetzung"