Is Translation Crowdsourcing Unethical?
The theme of the LRC conference in Limerick last month was “Localisation in the Cloud”, and unsurprisingly the issue of translation crowdsourcing was topical – in particular the “free” crowdsourcing that has been used by social networking sites like LinkedIn and Facebook. One conference attendee, a professional translator, raised the issue of ethics and this form of translation, arguing that it was possible that some 13-year old child from a third-world country could be doing work for a profit-oriented organisation.
Maybe this does raise an issue, though to be honest I think it’s somewhat moot. The software industry has been crowdsourcing like this for years. Just look at public betas. Or look at how complex software is written, from open-source software to WordPress plugins to Globalsight! Up to now I have not seen any LinkedIn groups called “Testers against Public Betas by Commercial Businesses” popping up. Neither has there been any ethical concerns about who is testing beta software “in the cloud”. I think the ethical issue is a red herring, thrown in to support an unwinnable argument.
The reality is that Translation Crowdsourcing is here to stay. It will compete with Machine Translation and professional Human Translation in the localisation ecosystem and it will find its niche. But it won’t take over that whole environment either. There is no solution on the horizon for avoiding human interaction in things like Life Science or Legal translations – the quality standard needed is just too high.
So where will Translation Crowdsourcing make an impact in the software industry? I think it will have a big impact in certain types of translation work:
- For “free”software (or at least where the end-user perceives the service as free), and where there are users who want to avail of that free sofware in their chosen language. See the Firefox web browser.
- For languages where there simply isn’t a business-case due to market size, and where there are users who are enthusiastic for their language to be supported. An example given at the LRC Conference was Albanian. Interestingly, someone who utilised crowdsourcing for translations mentioned to me that the model didn’t work so well for languages such as Dutch or French. French and Dutch speakers are pretty well served and don’t feel that they need to contibute to free translations. They can simply get an alternative that does support their needs.
- Where initial quality is not the top priority*. Scratch out anything to do with Life Sciences or Pharma! I say “initial” here, because there is nothing inherent in a well-run crowdsourcing system that precludes quality from evolving to be extremely high. Sure, there may be concerns about the quality provided by non-professional translators. But this can be mitigated by the size of the crowd as they cross-check each other’s work, or by adopting a hybrid model to involve a small number of professional linguists to review.
- Websites are good candidates for translation crowdsourcing. Non-professional translators need to see the results of their work quickly to maintain their enthusiasm for completing the work. For software (unless it has been architected with this in mind), this can be tough to do, but it is possible. The real-time nature of web applications enable people to see their inputs soon after they have been provided.
- Timelines are flexible. It is difficult to incentivise amateur translators to provide their inputs in a timeline that suits you, the project manager. If you have short, hard deadlines, then you really need to be waving money at motivated professionals.
- It’s not a one-off. It costs money to put in place a translation management system that supports crowdsourcing. The initial investment is likely in most cases to be higher than the cost of one-off professional translation services. For crowdsourcing to work, the ROI needs to be there for the long-term to justify the initial investment.
In short, I think the railing against Translation Crowdsourcing is understandable, but misplaced. There remains a good market for professional translators, and they need to focus their energies on ensuring that the promises about high-quality, timely translations are kept. If they can do that, there will be plenty of work to go around.
* The asterix is because professional translators have been very huffy about quality when discussing this issue. But the professional translation industry is not immune from poor quality. I have personally seen an instance where a well-known LSP supplied back appalling translation quality that failed quality tests for two-thirds of the languages they were contracted to provide. These guys are *supposed* to be using professional, trained translators. If they’re not, then professional translators should be campaigning against this potentially-fraudulent practice. Either way, the quality reputation of professional translators is not altogether proven, and it will be interesting to see if any hard statistics on quality come out in the future.
I like how you compared it to other crowdsourcing aspects of software. I don’t think any professional and confident translator should/would feel threatened by crowdsourcing. Crowdsourcing is not going to be a bigger threat to professional localization compared to what open source is to proprietary software.
There is no question to quality if the size of the crowd is large enough, like Facebook. The projects or businesses that can benefit from the current form of localization crowdsourcing, however, will greatly depend on the nature of the business.
Would I contribute to LinkedIn localization effort for Turkish? No. I don’t see why I would donate my time for free to a business which will generate revenue out of my free contribution. I do and will, however, contribute to others that I see as fair.
Will there be people contributing to LinkedIN localization effort for Turkish? Sure. Do I see this as a threat to my job or future? Not at all. When technology kills an opportunity, it creates dozens of others.
If you work in the technology field you need to keep up with this continuous change. If you don’t want to, you should probably switch careers as soon as you can cause not liking it doesn’t stop the change.
Congrats on the new blog. I will be posting about the new Facebook API for translation and will add the link to here in a few days…
Thanks for your thoughts, Emre. When you think about it, the conditions needed for a successful translation crowdsourcing are present in only a small percentage of cases. LinkedIn and Facebook will take the headlines, but for most for-profit software companies the model is not a good fit for their products or processes.
Or you can consider crowdsourced translation on another plane – not about getting something localized, but about engaging global customers/users further into the user experience. It’s hard to decontextualize the Facebook translation model from the other features offered to users as a competitive strategy in this regard.
@Emre, I totally agree with you that no professional and confident translator should feel threatened by crowdsourcing. I am only speaking for myself here, but I sincerely doubt that even a tiny fraction of crowdsourced translations would be capable of meeting the combination of tight deadlines and high quality (accuracy and a compelling writing style) my customers typically demand.
Crowdsourcing is here to stay, so rather than whining about it, find a niche where you have value to add as a professional translator (if you don’t have any added value over crowdsourced translation, it’s time to start asking yourself some important questions about your translation and marketing skills).
I do think it is a good thing and important on a symbolic level that ATA and other professional translators’ organizations take an official stance against for-profit enterprises using crowdsourcing, but mainly as an awareness-raising tactic. I don’t think they, or we as individual translators, should spend much more time or effort on the issue than that.
We’d all be better off spending that time finding those niches and learning how to position ourselves as valuable — no, essential — business assets to any multilingual project team.
[...] on Colin Cooper’s blog on localization and on the Naked Translations blog there have been [...]
I think it is worthwhile for all of us to dig into this to better understand where/if it makes sense for Buyers, LSPs and Translators.
Much of the criticism of crowdsourcing assumes a zero sum game or market. The thinking is, that if the crowd is involved, a professional translator is displaced and denied payment for services they would generally have rendered for payment. To some extent this may even be true in a world where source content grows by 10%-20% a year and where very standard localization source content is the only focus of the crowd effort like FB and LinkedIn which to my mind are NOT good examples of the real potential of this phenomenon.
However, I think as the web becomes much more global and corporate website is the first point of contact for information of high value to global customers, the picture changes.
I think the more likely scenario we face today is that the amount of source content growth is much more dramatic. It is quite possibly 10X or maybe even as much 100X the amount of source content that the localization industry translated last year. The demand for more information to be translated is growing exponentially.
So then how does a global company looking to grow in new markets solve this problem? Or increase content for existing global markets as community content grows in importance?
1. The global companies are not going to be given much more (if any at all) money to accomplish this
2. There are not enough translators to translate 10X or even 2X the source content let alone 100X which is where we are quickly heading
3. There is not enough time to use a standard TEP process even if money were no object
So these companies either accept that this cannot be done, or they try completely new approaches to address this need. I think that an emerging model to allow 10X content to be translated at high quality levels is developing and has the following characteristics:
• Increased use of MT (especially SMT that can learn and continuously improve)
• Increased use of translation technology platforms (not SDL) that enable large scale collaboration with robust tools and processes
• Much better interchange of linguistic data across platforms and products and vendors (which by definition rules out SDL)
• Initial guidance and steering of these “massive scale” projects by skilled language professionals
• Increasing involvement and use of the bilingual community (or crowd) to get large projects done especially in “new” markets
• The development of virtuous cycles that enable large amounts (millions of words) of content to be progressively translated at higher and higher quality
I think we are already seeing evidence of this with the Lionbridge “Translation Workspace” announcement, Lingotek has a crowd and translation management environment designed to help you use both professionals and volunteers and identify quality drivers, and Asia Online is introducing Language Studio Pro to enable the MT piece to be much more tightly integrated into this process: (www.languagestudio.com ). Moravia is one of the first LSPs to jump into this new model and I am sure we will see more enlightened LSPs also do so in future.
I think an emerging model for very large amounts of content will be Statistical MT + Expert Linguistic Steering + Professional Translation + Community or crowds
Can anybody else suggest other ways to address a 10X or a 100X growth in source content that begs to be translated at as low a cost as possible?
There are also very active discussion going on about this subject in LinkedIn in teh Localization Professionals group as well as the G11n – Globalization Professional groups
You’re missing the real threat, it’s using crowd-sourcing to edit results in Google translate that is really scary for the industry.
Care to elaborate on how this will cannibalise any more than a niche part of the market?
Finally! An analysis of crowdsourcing in translation that has a sensible, professional tone to it!
[...] How Will Crowdsourcing Change the Translation Industry? (Lionbridge)The Official Google Translate BlogIs Translation Crowdsourcing Unethical? [...]
Thanks for your thoughts, Emre. When you think about it, the conditions needed for a successful translation crowdsourcing are present in only a small percentage of cases. LinkedIn and Facebook will take the headlines, but for most for-profit software companies the model is not a good fit for their products or processes.