Machine translation obtained using Elia by Elhuyar
eu
Yearbook 2021 | Aitziber Agirre Ruiz de Arkaute (Elhuyar aldizkariaren zuzendaria)

Zientzia.eus is a website that collects all the scientific content of Elhuyar. It receives news and written reports, videos and audios since 2001 and has more than 23,000 publications. In recent years, in order to improve web accessibility and user experience, we have implemented a series of functionalities that apply artificial intelligence in language, such as the possibility of listening to news through speech synthesis, a multilingual search engine, the possibility of relating similar contents and offering an advanced reading of them.

A year ago, the automatic translator Elia (elia.eus), developed in Elhuyar, was implanted, and since then the written contents can be read automatically translated into five other languages (ES, FR, EN, CA, GL). For the time being, it only works on written content, although in future it is planned to introduce a technology for automatic translation of video and audio content.

The implementation of the translator in the drafting group of Elhuyar Science has not involved additional work or a change in work habits. At all times it has continued working in Basque and creating in Basque, the translations of content are generated automatically and their storage on the website is done entirely automatically. Storing all this information on the website has led Google to index these automatic translations into other languages and to show the translations themselves in the search results.

Quantitative and qualitative leap

The implantation in the museum of science has been a practical and significant example of the demonstration of the potentialities of Elía"s new technology. Immediately after the implementation of the automatic translator, visits to the website experienced a dramatic increase, up to almost three times the number of new users (from 111,790 new users to 306,036, representing an increase of 171%). It is undoubtedly the greatest boom in science in history. The total number of users has also increased by 164% and sessions 93% annually. That is, almost tripled users and almost doubled sessions.

Regarding the origin of the visits, until 2020 most users were Spanish (we have no way to discern the data of Hego Euskal Herria and Spain), 94%, and 0.7% of France (including Iparralde). However, following the implementation of the automatic translation, the number of Spanish and French users has increased dramatically (124% and 810% respectively), but users from outside these countries have risen from 5% to 18%. The website receives visits mainly from Mexico (5% of the total), USA, Colombia, Argentina, Peru, Chile and China.

That is, taking advantage of Elia, Elhuyar"s media has made a significant and qualitative quantitative leap. An important part of the Basque population that does not know Euskera has approached our media and brought readers from other countries closer to the Basque world, giving visibility to the Basque people through research and visions of the Basque Country. Moreover, many Basques that tend to search in Spanish have found in Google the content of science written in Basque language in science .eus and originally in Basque, something that until now we could not attract.

In fact, this extra traffic in science stems from searches in other languages. Now, for example, a reader searching in Spanish for information about the old scientific project 3MAG, news, consulting on Google “3 million African genomes”, will find directly the article of eus science in Spanish. You"ll get the National Geographic website first, and the science website second. Both in Spanish, although in reality the second is written in Basque.

The influence of the machine translation has been spectacular: now Google shows up to 8 times more science results. It has gone from 828,000 appearances in 2020 to 6,980,000 in 2021. Indexation has provided great visibility and dissemination to the website. Yes, five languages different from the Basque language. Therefore, in order for our objective to be met, it is essential that it be clearly stated in those contents that the original is written in Basque and that the form of referring it to the original content be added. This will give real visibility to the Basque Country and the contents of Basque. This may result in Euskaldunes used to conducting searches in other languages to the Basque media. We have new tools.

Another alternative translation model for content at the time of correspondence

The website of Elhuyar magazine has been otherwise integrated into the automatic translator. It was decided not to index in Google the erdal translations, so that the only way to read a content of the journal in other languages is to request the translation of the content on the journal"s own website. That is, the translations are done at the moment, they are not stored on the web. Thus, if a search is made in another Google language about a topic of science, the reader will not find the contents of the journal directly in Spanish, French, English or another language; the original Basque content of our website will be the only way to arrive at its translation.

In this model, the result corresponding to the number of visits was very different from that of the sciences. Also in Elhuyar magazine, initially, translations were saved on the website itself, indexed in Google, and the increase in visits was even greater than in the case of the science journal. When the model was changed and automatic translations were stopped, Google ignored the translated content and traffic quickly returned to the usual figures and characteristics.

Google’s indexing of web content stored in languages other than the Basque language is on the table and there are two models of automatic translator use. The debates, Elia is another way to publicize what happens in the scientific-technological environment of the Basque country. Another resource for the dissemination of Euskera.

On the other hand, it should be noted that the large technological platforms (Google, Facebook, Twitter…) already offer automatic translation. Taking into account your great technological capacity and your determined commitment to providing a multilingual service, whatever our will, we will receive and collect the translations of these contents. Therefore, if the Basques do not use our own means, we will receive the translations into the filter of the big technology platforms.

Aitziber Agirre Ruiz de Arkaute, director of Elhuyar magazine