A student project showed the potential of machine translation
Machine translation (MT) can help to preserve small languages such as Finnish. Some claim that it may even reduce the need to study foreign languages.
“A thorough understanding of a text still requires strong language skills. The need for high-quality language will not disappear, and language professionals will have work to do in the future also”, notes Sampo Savolainen, a student of English translation.
Savolainen was the project manager in a project in which students translated course descriptions of the School of Language, Translation and Literary Studies from Finnish into English. In their work, the students used a machine translation engine supporting the Finnish-English language pair.
“In the future, the radical increase in the use of machine translation could allow Finnish speakers to use their mother tongue in some everyday communication situations that have previously required foreign language skills. Machine translation is definitely not making the Finnish language disappear”, Savolainen reckons.
Even at its best, a translation engine can only produce raw translations, which then need to be edited by a language professional. Nevertheless, engines have indeed improved as translation tools.
Course instructor Mary Nurminen, university lecturer in English translation and interpreting, states that machine translation will eradicate neither translation as a profession nor the Finnish language.
Student Liisa Hattinen compares a translation engine to the spell checker of a word processor: it reveals typing errors but cannot flawlessly correct them.
“Translation engines will never reach perfection. The use of machine translation has certain types of restrictions, but it might actually become quite a common translation method for certain purposes”, Savolainen says.
Nurminen estimates that new translation tools can even create new job opportunities for translators. Due to new tools, translators will be able to translate more efficiently and might also be asked to translate content that previously would not have been considered for translation at all.
Business-academia collaboration supports small languages
Translation programmes have shown the most development in big language pairs. Business interests favour the most commonly used language combinations, such as English and Chinese or English and Spanish.
“When it comes to small languages like Finnish, machine translation is unlikely to develop without cooperation between universities and businesses”, Nurminen notes.
The idea of this course was born when a colleague of Nurminen noticed that many course descriptions of the School of Language, Translation and Literary Studies were not available in English. To resolve this shortcoming, the colleague suggested that Nurminen organise a course in which students would be assigned an authentic commission to translate the descriptions.
The course deployed a machine translation programme designed within an EU project several years ago. Since its development, the Finnish-English language version had remained largely unused – and thus undeveloped. Therefore, the Swedish company that owns the programme gladly agreed to let the students in Tampere test the programme within their translation project. This way, the company also got the chance to collect new user feedback for further improving their programme.
At the start of the project, the programme seemed rather clumsy. As the project proceeded, however, the engine learned to offer successful translation solutions.
“The programme worked well for certain things, as many of the Finnish course descriptions are written in a very similar manner. Using an engine saves quite a lot of time compared to translating everything from scratch”, Hattinen says.
The course descriptions comprised a total of 279 pages and included many recurring words and sentence structures.
MT works best on repetitive texts
Machine translation works best for translating content that includes a lot of repetition, such as user manuals. It is not even meant to help in translating literary texts.
Last year in the Amazon online bookshop, however, someone tried to sell machine translations of copyright-expired literary classics. The texts had been translated from English into Finnish using Google Translate. Among the victims of this offence was author Arthur Conan Doyle, the creator of Sherlock Holmes. It took a day and a half before the books were withdrawn from Amazon.
“Machine translation is not meant for these type of translations, which is a good thing. It is not a suitable method for all texts, nor is that the purpose”, Nurminen states.
Translators of literary texts may even see machine translation as a hindrance. The usefulness of the technology also depends on the language pair.
Nurminen compares the current situation to the time when translators first began to employ translation memories in their work. Translation memory programmes search databases of existing human translations and help the translator by offering possible previous translations of the same or similar text segments.
“Now the translation engine has become a similar tool. The main purpose is to enable translators to do their job more quickly.”
*Savolainen underlines that even when done with the aid of tools, translating is about communication.
“Communication is not about replacing one word with another; it is about conveying ideas. Although translation engines are improving, they are still a long way from mastering this skill.”
The prospects of machine translation are promising
The future of machine translation seems promising. Not only can engines produce raw output for human translators to work on, but even raw MT output as such has its uses.
For instance, producing a raw translation can help to estimate how much space a text would need in a given target language. Further, multinational and multilingual companies could speed up their internal communication by publishing internal reports as raw translations only. Carefully composed texts are not always necessary.
Engines are already able to produce better MT output than many people would expect. Even the Finnish-English language pair is beginning to work.
Machine translation can be used in real time too. One example is a customer service chat combined with MT, which actually serves its purpose well. Although the language produced is not perfect, a sufficient level of understanding can still be reached.
Nurminen estimates that in the next 10 to 20 years, people will get accustomed to using raw machine translations in certain situations and will learn not to expect perfect language everywhere.
New innovations are emerging. One idea is to build a speech recognition system into Skype to enable speakers of different languages, such as a Spanish and an English speaker, to discuss in real time using their own languages. A prototype already exists, but is yet to be adopted in general use.
The student team succeeded well
The project team consisted of ten translation students at the University of Tampere. As a team, the students succeeded in their project very well. To Nurminen, one of the goals of the project was to introduce the students to project work.
“Luckily, we had an amazing project manager. I just read through the students’ commentaries about the project, and no one complained about communication, even though that is exactly where projects usually go wrong. I stopped attending the team meetings midway, as Sampo seemed to have everything under control.”
To ensure effective communication, project manager Savolainen asked all team members to use a shared communication tool that enabled everyone to receive the same information at the same time. The outcome satisfied everyone.
“The Finnish source texts were written by several different university instructors, while this student project had a close-knit team of translators. As a result, I would even say that these English translations are much more consistent than the Finnish originals”, Nurminen says.
Nurminen promises that she can organise a similar course again next year. During the course, students could translate texts for some other schools of the University of Tampere that need to have their course descriptions in English.
This article is a human-produced translation of the article Pieni kieli hyötyy konekääntämisestä. For comparison, the original article was also fed through Google Translate; you can read the unedited machine translation here.
Text: Heikki Laurinolli
Photograph: Jonne Renvall
Translation: Camilla Kiviaho