There are many inspiring enterprise localization programs within the language services space, and they are very often at the forefront of news in our industry because of several reasons, ranging from automation and technology to quality assurance and vendor management.
Even though it is still in the process of maturing, there's one specific buy-side initiative that has long been under the radar but deserves special attention: Red Hat Training. Here's why.
1. AsciiDoc as the primary format for localization
Red Hat Training recently transitioned from DocBook XML to AsciiDoc as the primary authoring format for customer-facing training content. AsciiDoc is a lightweight authoring format from the Markdown family, loved by technical writers for its simplicity and versatility. However, it's often perceived as a localization nightmare. To ensure smooth localization from a technical perspective, Red Hat Training developed highly customized file filters for AsciiDoc and has thus pioneered the availability of functional AsciiDoc parsers in three market-leading TMS. An interesting bonus is that the AsciiDoc initiative is open source.
2. Chapter-based localization
Instead of localizing entire books or manuals, localization happens on a per-chapter basis. Individual chapters are translated while the English course content is still in development, with the goal of accelerating the time-to-market for translated courses once the English content is finalized. The result is a metadata-rich continuous process where high volumes of content are localized efficiently.
3. Tech stack driven by open-source software
In addition to AsciiDoc, Red Hat's tech stack is primarily characterized by cost-effective open-source software initiatives including GitHub for version management and Jenkins for building CI/CD pipelines.
4. Customized machine translation
Machine translation is an integral part of Red Hat's localization program, following an MT First strategy. MT is deployed as one of the primary linguistic resources, with all content processed in dedicated MTPE processes. Among the MT assets are custom-trained models and adaptive models.
5. Proprietary QA mechanism
Quality assurance is probably one of the most obsolete disciplines in localization due to its rule-based nature. QA algorithms struggle to identify "real" translation errors, leading to high amounts of false positives. Red Hat Training has developed a target-only QA algorithm that scans translated AsciiDoc files for technical issues preventing correct output generation. The algorithm generates HTML logs that are shared with relevant stakeholders in a fully automated way.
6. Custom automation
Acknowledging the limitations of commercial automations (e.g., TMS connectors, QA), Red Hat Training embarked on in-house development of every automation initiative. The different components in the tech stack are tightly glued together through a set of scripts automating every step in the pipeline, from source content retrieval to preprocessing, API connections with the TMS, post-processing, QA, and publication. The different automation assets are accessible through a custom UI for human intervention, and can always be adapted and enhanced without third-party dependencies.
Finally, it must be mentioned that the Red Hat Training Curriculum program is excellently managed by Lène Spee in the position of Localization Manager.
At C-Jay International, we feel privileged to be a part of this fascinating initiative and we look forward to further developing the program.