Should your business rely on open source localization tools?
There exists an assumption that if a translator doesn’t have a paid CAT tool license, he or she cannot be trusted as a professional. Who thinks that? Why, fellow translators.
But can we really judge a professional by their toolkit or by the money they spend on special software licenses?
More importantly, what are the alternative technology solutions available in the localization industry?
The open source approach
Take a middle-sized IT company and ask them about their localization demands. Quite often the first area they investigate will be open source localization. Bigger companies, such as Adobe or Microsoft, claim that they “believe in open source” – when speaking about localization.
In fact, Microsoft ranks as the world’s top open source contributor, as measured by the sheer number of employees active on GitHub. Chinese companies like Alibaba, Tencent, and Baidu are also significant contributors.
In modern IT, open source conditions the innovations in AI, mobile, big data, and machine learning (ML). TensorFlow, an open source software library for ML that enables software developers to apply deep learning to their products, has shaped the ways in which ML has been developed.
A neural network playground which TensorFlow presents is serving the purposes of content reuse in technical documentation as well as other language-related tasks. For example, the Google AI team open-sourced Lingvo, a framework for building neural networks in TensorFlow, for the tasks of machine translation, speech recognition, and speech synthesis. There are also other deep learning libraries that are being used in the translation and localization industry (e.g. MXNet, PyTorch).
The future of enterprise technology is increasingly driven by open source. Let’s have a closer look at how the localization industry is responding to this trend.
The pros of open source localization…
Software (SW) developers generally do not like to use someone else’s code. At the same time, many agree that open source software (OSS) serves as a model of open, professional collaboration.
There are various types of OSS licenses and two basic ways of how you can actually implement them in your business:
- Either you take the open code and further develop it in-house using community solutions, or
- You pay for an enterprise version, which offers business-level support
In both cases, while the software itself may not exactly meet your unique requirements, its open source structure allows tailoring the solution to your needs within an affordable development time, while you maintain full control of the direction the software is heading.
Here is why open source localization can work:
- You don’t have to pay third-party companies. Also, you’re saving yourself from a situation when you did pay – and purchased a CAT tool license – but received poor customer service or support.
- Open source means flexibility which cannot be guaranteed by a third-party, proprietary tool. With open source, you can develop a tool further as per your own understanding of the desired infrastructure, functionality, and preferences.
- Open source also provides more stability. Users relying on publicly distributed open software can be sure that these tools, as well as their support, won’t disappear if the original creators stop working on it.
- You can fork the open source solution, secure it from the public, combine with other solutions and build your own localization SW on top of that. But it depends on the type of license. Some open source licenses (e.g., those which are called “copyleft”) say that anyone who releases a modified open source SW must also release the source code for that SW.
- With open source, you host your localization services and store data in-house, which promises better data privacy.
- The speed of development is also comparatively higher: different programmers can simultaneously fix, update, and upgrade open source software more quickly than proprietary software developers release features.
Understanding what’s out there – and what you can do with it
As mentioned above, in the majority of cases an open source localization tool won’t guarantee a perfect match for your desired functionality. Some may have:
- A strong component for a crowd-based translation (Pontoon, Zanata)
- A powerful editing/TM component (e.g. a desktop translation tool OmegaT)
- A comprehensive translation management system – like Globalsight, a project acquired by Welocalize
- NMT framework Systran and Harvard University built together (Open NMT) – now also managed by Ubiqus
- Or file format converters for translation processes (Okapi)
It is exactly this incompleteness of such solutions that opens an opportunity to play with them further.
That’s what happened with Serge, a tool created in Evernote to help localize products and marketing materials continuously into 25 languages. Evernote uses a CAT tool named Zing, which is a forked version of Pootle, a free, open-source, web-based translation server that supports working with both community and paid translators.
Another successful open source story comes from the project Translate5, which started about 15 years ago with a thought, “Why are there no good open source tools for localization?” and a desire to change that – while actual development started in 2011, when a client demand for a web-based review solution appeared. It gave an impulse for the project, which further evolved in 2015-2017 from an online review system to a cloud-based open source translation system. This evolution from a review to translation system was crowdfunded in three phases. The team now has five people who concentrate on actual development.
Opening the source code for the public presents a way to check if and how the community adopts the idea. That’s why at Github, those interested in open source localization can certainly find things to explore. Whether it is the source code of a term extractor or a way to sign up for the Microsoft Open Source Localization Program and directly contribute to the localization of open source products by Microsoft, there are quite a lot of possibilities. Of the most popular ones, Matecat should be mentioned (more than 16K commits). They have a specific Lesser General Public License, though, which presents some challenges.
…and the cons
A special kind of “free”
A common misconception about open source is that it is totally free of charge. However, it’s not. The term “free” does not refer to monetary cost, but rather to the freedom of using the software and source code. In fact, open source SW creators can and frequently do charge money for an enterprise version of the support or even help with deployment. Some programmers believe that charging for services and support rather than for the software license is more rewarding. The software stays free, but they can still make money by helping others use it to the best possible extent.
Is open source the recipe for you?
There is a concept that open source is like sharing a recipe. At first, you are given a recipe, which you improve. Then you serve the changes up back to others. However, like with any recipe, something may go wrong. Whether it’s a poor “cook,” or a missing “local ingredient,” you can’t really guarantee a 100 percent result. Sometimes it is easier, and makes more sense, to go to the shop and buy a ready-made plate in fancy packaging.
Open source stands for community
Open source is, by nature, community-driven. However, the localization community is not necessarily interesting to the broader SW community. This means there is lower participation and help from the crowd when it comes to troubleshooting, continuous improvements, and quality advancements.
Community means not only consuming but giving back to it, too. Maybe you’re not ready for it, or don’t plan to invest time in the actual giving back. However, there may come a time “when you need to make some changes because otherwise your products won’t work well and you won’t have credibility – people won’t listen to you and then you’re stuck with supporting customers of a code base you have no say in,” as stated by Raghu Ramakrishnan, Microsoft CTO of Data.
Last but not least, other companies might begin to use your open source localization project to their own benefit, which at some point may outgrow your own. At the same time, IT majors have learned to “open source the competition.”
OSS is found in nearly every area of today’s IT industry, proving to be beneficial for enterprise, non-profit, and personal use. In the bigger IT industry, open source projects are being improved by businesses and individual users contributing back.
Localization software, however, is a smaller niche. Though we have more than 450 localization technology solutions on the market, less than four percent of them is open source. Traditional software licensing prevails in localization tech. The size of the localization tech niche is about USD 800 million, and it’s not promoting itself very well.
Marketing is one of the two main challenges with driving an open source project: “As an Open Source company we do not have a big marketing budget like proprietary tools. We are very focused on development and service on users,” says Marc Mittag, the author of the translate5 project. Coordination of funding of features between different companies in the community is another top challenge, according to Marc, “which is also a big plus, because it brings you close to the interests and contacts of your users.”
There’s no critical mass of the localization tech users who need or could really drive the OSS development – not just yet. But it might change. Look at the NMT side of the business, for instance, and you’ll notice that more and more open source projects continue to emerge in this area.
An opportunity to seize:
- For users, OSS presents a way to learn and try new approaches with lower costs and less time. Localization buyers, as well as LSPs and localization tech providers, can try combinations of OSS coming from outside of the industry, and working together to make better software. This will increase the number of alternative solutions available in the localization industry and, in the long run, may even influence the pricing politics of the available third-party tools.
- For OSS creators, in addition to selling consulting and support for open source products, they may sell exemptions to the terms that the SW is licensed under. Donations and crowdfunding can also do the trick.
In an industry that very well realizes how important and valuable free communication is, there must be more space for free collaboration. Open source might be the avenue to achieve it.
Stay up to date as Nimdzi publishes new insights. We will keep you posted as each new report is published so that you are sure not to miss anything.