Is Machine Translation a solution?

“Post-Editing Free Machine Translation: From a Language Vendor’s Perspective” 

 by Luciana Ramos

Who dares deny humanity is in constant change? Changes have always driven evolution and evolution is only possible as long as innovations are embraced by users, consumers…people at large.

After many years of development and research, machine translation has taken the floor and it’s worth listening to what it has to say, or taking advantage of what it has to offer – why should we say MT is not worth using?  If so, we will be denying all the hard work technology experts and researchers have been doing for more than fifty years: in 1952, the first MT conference was held at MIT (Massachusetts Institute of Technology). At that time, full automation of good quality translation was considered as virtually impossible and human intervention either before or after computer processes was thought to be essential (Hutchins, 2001).
Since then, much has been done not only on the developers and researchers’ side, but also from the translators’ community, and it is time to expect for the massive use of Machine Translation solutions.
1 – Why Choosing Machine Translation as Part of the Localization/Translation Process

Nowadays, MT technology and practices are varied. Some software developers or content   creators (such as Microsoft, IBM, just to name a few) have built their own systems and outlined their own guidelines to implement it. The general public has Google Translator Toolkit; and many language and technology companies have developed their own customizable engines (such as Asia Online’s Statistical Machine Translation platform, Welocalize’s GlobalSight Technology, Systran’s Hybrid Machine Translation Engine, among others) to cater for a wide array of content types and fulfill particular needs.

By and large, and in controlled-enough scenarios, machine translation helps clients reach foreign-language markets in shorter times and to a larger scale. But, what is a controlled scenario? And what does it take?

It encompasses a set of rules and procedures regulating and sensing MT post-editing; it takes, above all, experienced professionals in the technology and language areas, so that reasonable and adequate quality goals are set, procedures are delineated accordingly, and pre and post-editing phases are included in the loop. Under such circumstances, MT can offer an effective shortcut to multilingual communication.

2 – Tailoring Content for MT and Tailoring MT to Content

Once the decision of using MT is made, a linguist or translation expert is essential, either to make content suitable for MT or to define the best MT-based approach based on content and intended use. As a start point, content type and language used in the source text should be taken into account to choose between statistical, rule-based or hybrid machine translation. Text intention is also relevant for making the decision, even on whether machine translation is feasible or what type of post-editing will be needed. Linguists and developers working together will render the best outcome in terms of time and cost efficiencies, considering the translator’s input as the most useful device to filter errors out, thus avoiding the need of repetitive corrections or enhancing the engine.

The role of the human translation editor and MT post-editor should be clearly differentiated. Post-editors have to be able to focus on adjusting the MT output so that it reflects and conveys the meaning of the source text as accurately as possible (Krings, 2001). On the contrary, human translation editors tend to strive to disguise the fact that the text has been translated (Senez, 1998).

3 – Google Translator Toolkit – Free Machine Translation Engine

Both for the individual and commercial user, this product from Google offers a free, easy to access and use of global engine for machine translation. As a language service provider, we, at Ocean Translations S.R.L., have been using this engine as the initial translation step for several clients (only when specifically requested) and results were encouraging. Our linguists, in their role of post-editors, are satisfied with the quality they produce and the fees they get for it. It takes them one hour to post-edit 500-800 words (versus 800-1000 words edited when translation was done by another translator). A sustained flow of work (related mainly to the lengthy nature of such assignments) is also reported as one of the advantages of working as post-editors. As an average, around 650 words are post-edited per hour, and an eight-hour working day is possible under post-editing working conditions, with minimal breaks in between, such as when working with human translation.
As an illustrative fact, we have recently completed a job of more than 90,000 words post-edited and revised (English into Spanish) in 9 calendar days (from client’s confirmation to delivery) involving 1 project manager, 4 post-editors (one designated as a lead post-editor) and 2 reviewers, and estimated cost savings for the client were approximately 40%.

As requested by one of our main and most valued translation clients, the Department of Reproductive Health and Research (RHR) at WHO Headquarters, we have carried out an analysis of shifting from the current model (CAT Tool-based) to Google Translation Toolkit for the localization of the Reproductive Health Library, for which updates are generated every year. Our report to them clearly reflects our overall experience in relation to this engine, and it is encouraging –although challenging, of course– to see how happy both parties can be with this solution, bringing time and money savings to one side, and a better utilization of resources and contracts for larger volume to the other.

4 – Conclusion

As in any environment fear to change has to be overcome and collaborative alliances are the best tool for this. Machine Translation is no longer “the future” and it is here to stay. After many years of research and development, it has become a useful tool available to the population at large, which will yield the best results when used by the best trained and most qualified professionals: translators, terminologists and software developers. Any translation or localization endeavor has to be duly discussed and a pros and cons analysis about using MT should be performed by a team of language and technology professionals.

In a nutshell, adding Machine Translation to the production cycle is an undeniable advantage, especially for large volume contents. Leveraging CAT and MT technologies with human post-editors, the translation industry may experience a significant increase in its figures and it does not have to mean human professionals of the language industry will lose their jobs – on the contrary, more and longer working opportunities may arise, as we have experienced, as translation becomes an easier and more flexible service to buy.


Abridged version of “Post-Editing Free Machine Translation: From a Language Vendor’s Perspective” – presented at the AMTA 2010: the Ninth conference of the Association for Machine Translation in the Americas, Denver, Colorado, October 31 – November 4, 2010 – Original Version can be found here

About Luciana Ramos

Graduated as a Technical-Scientific and Literary Translator and Interpreter from a well-known institution based in Rosario, Argentina, Luciana started working as a freelance translator for several international firms until she became part of permanent staff for an American company as a member of a virtual office. From then on and as a standalone professional again, she has translated, edited and proofed contents for the English-Spanish language pair, and training new staff on the go to meet the increasing demands of linguistic services. Later, as one of the owners of Ocean Translations S.R.L., she has gradually devoted to staff recruiting and training, always leveraging and deploying state-of-the-art technologies. Merging her training and recruiting roles, she has delivered several lectures both in Argentina and abroad. She left Ocean Translations in 2010.

About Ocean Translations

Ocean Translations is a global provider of high quality communications solutions, delivering expert language services across a broad range of sectors. Their professional services consist of the following: Translation, Localization (L10N), Internationalization (I18N), Interpreting (simultaneous, consecutive and tele-interpreting), Transcription – hard copies and audio files, Multilingual DTP and Language and Cultural Training. Combining sound management and tracking ethics with linguistic teams specialized in both cultural localization and subject matter; Ocean Translations delivers fluent and effective language solutions (English, Spanish, Portuguese and Quechua) to serve a wide range of fields. As an ISO 9001:2008 certified organization, Ocean Translations teams follow rigorous processes in order to produce High-quality, culturally accurate and timely-delivered translations.

Share this post


Get our best content straight into your inbox

Latest blog posts