What is crowdsourced translation?

What is crowdsourcing?

Crowdsourcing has been defined in a number of ways, but at its core, it can be described as a method of problem-solving in which the combined knowledge of as many participants as possible is used to come up with the best solution to an issue. Generally, the procurer publishes a call for volunteers who then begin to solve the problem in the format established by the organiser. The problem-solvers are mainly rewarded for their efforts with intellectual satisfaction. Probably the best known and most popular crowdsourcing project is Wikipedia – a multilingual online encyclopaedia which is freely accessible and can be supplemented and edited by anyone.

Such outsourcing projects are usually launched through a website that acts a marketplace, where procurers can publish a request for services and interested parties can then post their offers. The same principles have also been applied to translation, yielding so-called crowdsourced translation, wherein a community of, generally, non-experts collaborates on creating translations on, generally, a volunteer basis.

How exactly does crowdsourced translation work?

As in other instances of crowdsourcing, the idea arose out of a need: to translate a massive amount of information quickly and with high quality. As neither professional translators nor machine translation engines can do both at the same time – translators produce high-quality translations, but not fast enough, while machine translation engines produce instantaneous translations, but with poor quality – a collaborative method of problem-solving was developed: crowdsourced translation. In the case of crowdsourced translation, translations are produced through the simultaneous efforts of a large number of translators, sometimes working as volunteers, sometimes remunerated for their work. Some of the translators may be professional language service providers trained in translation. Others, meanwhile, may be amateur translators.

Similar translation models exist under a variety of names, including community translation, hive translation, collaborative translation, social translation, and CT3, or community, crowdsourced, and collaborative translation – all of which essentially employ the same method and differ only in a few minor details.

In general, crowdsourced translation can be divided into two categories based on the profile of the translators involved: non-professional translators – e.g., Google Translate users who improve the translations of the machine translation engine – and professional translators, who use various crowdsourced translation service platforms (e.g., TM-Town, Gengo, Smartling, etc.).

As a result of the rapid development of existing information technologies and the creation of new technologies, translation services are becoming increasingly accessible to everyone, and a number of translation project models involving volunteer translators have emerged. Here are the three most popular ones.

Machine translation editing

Machine learning is currently one of the hottest trends in the world of translation. Although it has been suggested that artificial intelligence based machine translation engines may soon perform as well as the best human translators, the truth is that there are still a number of obstacles to overcome before that will happen. In addition to years of coding and developing sophisticated algorithms, training these engines requires a large number of people to review the accuracy of the input source texts and their translations. While machine learning may well be the future of translation, translation engines based on machine learning are as yet still too expensive and labour-intensive for the average company.

Community translation

In contrast, community translation, in which the aid of members of the relevant community or group is enlisted to translate the necessary content, is a much more affordable and simpler solution. For example, the online gaming communities of Second Life and Steam have excellent community translators, who help translate their content into dozens of languages. However, while affordable, community translation may not be suitable for translating more demanding content, as the quality may not always be up to par.

Crowdsourced translation

Crowdsourced translation similarly involves a large number of translators, but compared to the previous two translation models, crowdsourced translation projects are generally better managed and yield higher-quality translations. This is because these projects and the translators involved are generally managed by language service providers, who monitor the feedback given by customers on the translations in order to gather information on which translators are qualified to translate, for example, technical texts, and are thus able to ensure high translation quality even in the case of more demanding content. The language service providers also offer easy-to-use platforms that enable the customer to communicate directly with the translators. This eliminates the middle-man, making the translation process simpler and thereby saving both time and money.

However, in addition to the aforementioned need to translate a massive amount of information quickly and with high quality, economic aspects, of course, also come into play: In today’s highly competitive environment, everyone is trying to keep their costs as low as possible. As crowdsourcing largely relies on volunteers, it may seem like a great way to minimise costs – at least at first glance. But how does it hold up in practice?

Advantages of crowdsourced translation

Translations are primarily evaluated based on three criteria: quality, speed, and price. High speed and low price are the main benefits and reasons for why crowdsourced translation is gaining popularity.

1. Translation speed

Crowdsourced translation involves many people working concurrently on the same translation project and correcting each other’s mistakes, which makes it possible to save time on editing the translations and enables crowdsourced projects to be completed in a relatively short amount of time.

2. Translation price

With an existing community of translators, the necessary platform, and skilled project management, crowdsourced translation is also affordable.

3. Variety of language pairs

Crowdsourced translation allows bringing onboard a large number of speakers of many different languages, and there are essentially no barriers to how many languages or which languages the projects can involve. The challenge, however, lies in finding suitable translators and motivating them to contribute.

Disadvantages of crowdsourced translation

Launching, managing, and ensuring the quality of crowdsourced translation projects is not always easy. Not only that, but often it may not actually be much cheaper than engaging the services of a translation agency.

1. Quality may not be up to snuff

The biggest problem with crowdsourced translation is certainly the quality of the translations, which may not always meet expectations. The volunteers offering their services in crowdsourced translation projects are rarely professional translators, linguists, or subject matter experts.

2. Finding translators can be difficult

High translation quality can also be difficult to achieve if you cannot find enough speakers of the required languages – the result is too few translations. The lack of suitable translators is a problem especially in the case of small languages, such as Estonian.

3. Inconsistent translations

A single word or sentence can often be translated in a number of ways, which is why crowdsourced translations are frequently very inconsistent and may need considerable editing to iron out the inconsistencies.

4. Hidden costs

While, at first glance, crowdsourced translation can appear inexpensive (after all, you aren’t paying the translators), it still involves a number of costs, including the costs of setting up a common platform; recruiting, constantly communicating with, and motivating the translators; managing the project; and checking and editing the translations. In the case of more exacting translation projects, the crowdsourced translation may even need to be ordered from a translation agency as a service.

5. Confidentiality issues

Confidentiality is difficult to ensure in a crowdsourced translation project, as there will be many participants, who are generally not contractually bound.

Despite these disadvantages, however, crowdsourced translation still be beneficial in many cases. For example, it is always important for a company entering a new market to communicate with their very first customers in that country. Engaging these customers in crowdsourced translation projects and creating a community around yourself can be an excellent way to gain a foothold in the market. However, as already mentioned: in order for these projects to produce high-quality translations in a timely manner, they must be managed and there must be constant communication with the translators.

Which types of texts are suited for crowdsourced translation? 

Crowdsourced translation has been used relatively successfully for translating user interfaces (UI). Crowdsourced UI translation projects are often successful because the translators are mostly people who are motivated to use the software in their native language.

Crowdsourced translation can also be used successfully to translate user support, help centre, and FAQ (frequently asked questions) content, provided that the product or service is not too complex. If your content is highly complex or technical, you will probably need the help of a professional translation agency.

Which types of texts are NOT suited for crowdsourced translation?

Crowdsourced translation may not be the best option for translating marketing materials, as this requires a good understanding of your company, industry, technology, and the terminology of your particular field. In addition, mottos and advertising slogans are often impossible to translate literally, which can be a real stumbling block for volunteers in a crowdsourced translation project who do not know the background. Nevertheless, their help can still be a valuable way to get feedback and ideas on how to change your slogan so it would work better in the target language and market.

The translation of legal texts (including, for example, terms of service), meanwhile, should always be left to a professional translation agency. This is because such texts often involve much more than simple translation, as they may also need to be brought into line with local laws and jurisprudence.

Confidential material, which must under no circumstances fall in the hands of your competitors or become public, is also a poor fit for crowdsourced translation, as the latter makes it difficult to protect sensitive information.

What are the best crowdsourced translation platforms?

The most popular crowdsourcing platforms are CrowdFlower, Amazon Mechanical Turk, and Clickworker. CrowdFlower is a crowdsourcing and data mining company, which also offers, for example, integrated machine learning and cloud solutions in addition to crowdsourcing networks. The Amazon Mechanical Turk, owned by – you guessed it – Amazon, allows you to hire remotely located people to perform ‘microtasks’. Clickworker can be used as a free self-service platform or as a complete paid service. A more detailed comparison of these crowdsourcing platforms can be found here.

In addition, there are number of other service providers on the internet, including: Crowdin, CSOFT, Gengo, Localize.js, OneSky, POEditor, Transifex, Unbabel, XTM Cloud, and more. Which of these or the many other services available is right for you depends on your particular needs and means.

Who uses crowdsourced translation?

Finally, here are some examples of well-known companies other than Google who have successfully engaged volunteer translators and communities in their translation projects:

• Cross-Lingual Wiki Engine – for translating content;
• Facebook – for translating their website and UI;
• Twitter – for translating their UI as well as other people’s tweets;
• Microsoft – for editing machine translations;
• TED – for subtitling video recordings of TED Talks.

In summary, crowdsourced translation has both its pros and cons; it is not yet suitable for translating all types of texts, and to achieve good results, the company employing the method still needs to be willing to invest some of their resources: both time and money. Nevertheless, if properly executed, there is more to gain than lose. In the case of more important materials, however, crowdsourced translations should always be thoroughly reviewed and edited. Or simply entrusted to a professional translation partner.

Sources:

https://www.accelingo.com/crowdsourced-translation/

https://www.transifex.com/blog/2016/crowdsourcing-101/

https://blog.csoftintl.com/3-models-crowdsourced-translation/

https://www.language-translation-help.com/crowdsourcing.html

https://en.wikipedia.org/wiki/Crowdsourcing_as_Human-Machine_Translation https://ssix-project.eu/comparing-three-of-the-most-popular-crowdsourcing-platforms/

https://et.wikipedia.org/wiki/%C3%9Chisloome

https://pdfs.semanticscholar.org/b6fe/203b63e940e0c0014eb707c882a689df697d.pdf https://www.yumpu.com/xx/document/read/23974392/endc-proceedings-16-2012-kaitsevae-ahendatud-/80

https://slator.com/technology/the-elusive-uberization-of-translation/

https://www.vertaalt.nu/blog/dangers-crowdsourcing-translations/

https://www.morningtrans.com/crowdsourced-translation-does-it-work/