Specific topics discussed at the Chicago OSCAR meeting included:

• Translation Memory Exchange (TMX), especially certification and segmentation,
• word counting standards,
• Term Base Exchange (TBX) and OLIF developments
• related standards efforts (XLIFF and CGS), and
• the proposed OSCAR by-laws.

Each of these five items has significance for all groups and industries which perform or purchase localization.

TMX Certification and Segmentation

Mark Eckes of the Language Technology Research Center (LTRC) attended the meeting to discuss how software vendors should display the TMX logo that indicates their product has been certified to be in compliance with the TMX specification. Currently LTRC is the only organization recognized by OSCAR to certify vendor product compliance with the TMX standard.

According to the TMX specification, there are three levels of TMX compliance. As agreed at this meeting, explanatory phrases will be attached to the logo to characterize the levels of compliance. Further explanations will be made available on the LISA website. With this certification program in place, the level of compliance for a particular tool or application will now be independently verified and so will lend greater credibility to the standard, the industry, and the products.

As TMX has been implemented and used over the past year, it has been observed that importation of TMX data does not always lead to 100% capture of the exported data. Problems arise as a result of different text segmentation schemes in different vendor implementations of the standard. In particular, if text segments differ as a result of additional white spaces or punctuation formatting, a match will not be made. This reduction in the percentage of perfect matches results in increased manual checking and therefore higher costs. In fact, this problem of a less-than-expected percentage of perfect matches occurs even when TMX is not in the equation. It can occur just going between different tools in a suite from a single software vendor.

After some discussion, it was concluded that OSCAR should step forward and find a solution to this problem by defining an industry standard segmentation procedure. Once this procedure has been defined by OSCAR, in consultation with the various tools vendors, it can be integrated into the various tools in order to get closer to the expected percentage of expected matches. This initiative has been spearheaded by Yves Savourel, working in partnership with other LISA companies.

Word Counting

Differences in segmentation are not the only inconsistencies that need to be addressed to provide standards for localization and globalization. As most of us are aware, counting the words in a file, be it a source text or target text, using different tools can result in radically different word counts. When billing is done by word count, this can clearly result in a significant difference in the cost of a translation. Given the apparent magnitude of this issue, OSCAR voted to add to the job of the segmentation working group the related linguistic task of defining a standardized word counting procedure. Of course, both segmentation and word count procedures are somewhat language-specific; therefore, there will be separate procedures for various groups of languages.

TBX and OLIF

Thanks to the concerted efforts of the SALT project, a new draft of the TBX specification was submitted to OSCAR for review. TBX will do for terminology databases what TMX has done for translation memory databases. In addition to the specification, SALT is providing OSCAR with a TBX validating editor that will facilitate the process of determining whether a particular TBX file conforms to the TBX core structure and the allowed types and values of data fields.

In addition, there is on-going cooperation with the OLIF consortium aimed at facilitating the synchronization of terminology in the human translation and machine translation operations of an organization. Progress in this area will be reported in September at the world machine translation summit in Spain.

XLIFF and CGS

In addition to the various standards efforts that fall directly under OSCAR, there are several related efforts, including XLIFF and CGS. Yves Savourel, technical chair of OSCAR and also part of the XLIFF technical team, reported on how XLIFF is used in software localization and how it complements both TMX and TBX.

Jorden Woods, who hopes to bring his CGS effort into OSCAR, gave an introduction to how CGS will fit in with XLIFF, TMX, and TBX to coordinate global, regional, and local aspects of localization. CGS aims to provide the crucial information within a workflow about which units of text are to be translated, relations between translated units of content, and how the localization process should be carried out.

OSCAR By-laws

A draft set of by-laws for OSCAR was presented for discussion. These by-laws will formalize working and voting procedures within OSCAR and replace the informal procedures that have been used until now. This is a normal step as OSCAR becomes better known in the language industries and establishes more formal relations with other standards bodies.

OSCAR is growing up and contributing to society. Well, as least to the localization industry. But remember, OSCAR R US! Your continued support of OSCAR is appreciated.

OSCAR Steering Committee members in attendance in Chicago were: Alan K. Melby (BYU, OSCAR coordinator), Yves Savourel (RWS, OSCAR technical chair), Matthias Zeitler (Star), Steve Westover (SDL), Etienne Kroger (WeLocalize), Kara Warburton (IBM), Paulo Vanni (JDEdwards), Gérard Cattin des Bois (Industry Professional). The Administrative Chair in attendance in Chicago was Jorden Woods (GlobalSight). Invited guests included Mark Eckes (Language Technology Research Center) and Alex Lam (LISA webmaster).

The next OSCAR meeting will be held in Vienna, Austria, in November in conjunction with the Vienna LISA Forum.