Our Analysts' Insights

Multilingual knowledge bases (MLKBs) are an essential component for many organizations in today’s global economy. Many companies still try to adapt their legacy word-based repository into searchable documents online. However, this is not enough to meet the agile-ized, mobile-ized, and personalized requirements of their customers and prospects. These forces, along with current strides in AI, are blowing up traditional monolingual models for digital product content delivery, which will in turn affect multilingual creation and delivery models for knowledge bases.

CSA Research is currently interviewing enterprise buyers of language services about the challenges they face with MLKBs, how they’re resolving them, and what they’re preparing for in the future. Here are some of the preliminary results:

(If you’re interested in being interviewed, contact rebecca@csa-research.com.)

• Knowledge or customer relationship management platforms generally don’t support optimized localization. Similar to content management systems, the platforms used to create, manage, and publish KBs rarely support multilingual versions to the extent they should. Part of the reason is that organizations often customize these systems to the point that the out-of-the-box localized UI no longer supports their needs. Another sore point is search. It is often weak for languages other than English and non-existent for handling more than two languages at a time.


• Customers may be more accepting of machine translation (MT) than you expect.Participants in our interviews emphasize the importance of finding ways to engage directly with customers to confirm MT preferences. Google Translate – and now neural MT – have gone a long way to improving acceptance of both raw and post-edited machine translation. Interviewees share that even people in markets such as Japan, with its stringent quality requirements, are often open to machine-translated knowledge bases if it means they can gain access to content that would otherwise remain inaccessible.


• It’s time to investigate chatbots. Even if your company hasn’t begun serious research in this area, don’t be shy about kicking off informal discussion with colleagues in support, technical documentation, and product development. Conversations take different paths based on culture, so it’s critical for chatbot design to be world-ready from day one. For example, requesting personal information in markets such as China, Japan, and Korea must be done in different ways and at different points within a conversation than in English.


• Google isn’t the only game in town for search. Do you know how your customers in other markets enter your knowledge base? Unless they almost always access it from within your site, then you must make provision for SEO in multiple languages through Google – and whatever other search engines are common in local markets, e.g., via WeChat on mobile phones.


• Videos are often embedded in knowledge base articles and available through YouTube channels. If that’s the case for your company, then you must be ready to make the decision as to which ones – if any – you will localize. Though cheaper for production, subtitles may not always be the optimum solution if they cover too much of the screen for how-to type videos or are illegible on mobile screens.


• Migration of legacy content and platforms. This area continues to elicit the most groans – and the most advice – from our interviewees. No matter how much you prepare, this task always takes longer than what you project. Some interviewees recommend that you clean out older content prior to migration, while others prefer to wait until the new system is available. Others caution that automated scripts may not be as much help as you think since they require a fair amount of testing and never entirely obviate manual work.

Our research is confirming that companies should reassess the user experience for MLKBs to allow their customers to find answers quickly – regardless of who designs and creates the content or the delivery platform that provisions it. As a result, it may not be enough for localization teams to simply follow the lead of their monolingual KB counterparts. Rather, they need to find a seat in the teams that are re-thinking KBs – or push for such teams to be set up and get to work.