The National Relay Station, also known as Stáisiún Athsheachadta Náisiúnta na hÉireann, was created through the European Language Resources Infrastructure (ELRI Project) by researchers in the ADAPT Centre, which is located in Dublin City University. The ADAPT centre is ‘the world-leading SFI Research Centre for AI-Driven Digital Content Technology”. The ELRI project ran from 2017-2019 and was co-funded by the European Commission under the Connecting Europe Facility (CEF) programme. The aim of ELRI was to develop robust infrastructures for gathering, preparing and providing language resources to improve Machine Translation engines (in particular the ones of the CEF eTranslation service).
The National Relay Station was the first central repository in Ireland to share bilingual data on a national level and is currently being funded by the Department of Tourism, Culture, Arts, Gaeltacht, Sport and Media until December 2022. The project recently received over €200,000 from the Irish government. The NRS is used by public sector employees working with the Irish language, for uploading and sharing Irish language data.
An Irish eTranslation campaign was launched by The National Relay Station’s Research Assistant Jane O’Connor and Dr Teresa Lynn (ELRC National Anchor Point) in order to increase the number of eTranslation users in Ireland and gather feedback about the Irish language eTranslation engine. It was also a way of conveying to Irish stakeholders the importance of collecting Irish language data through the NRS, so that the Irish eTranslation engine can continue to be trained and improved.
The campaign was launched on the 27th of October 2021 by email to NRS newsletter subscribers and it was shared on the NRS Twitter, Instagram and LinkedIn platforms
Between the 27th of October and the 30th of November 2021, 6 reviews were provided from those working in the public sector including the NRS’s Research Assistant. Two reviews were provided from Irish Language Officers, two were from public sector workers working with the Irish language, and two were provided by researchers involved in Irish language research.
Participants were asked to give feedback by means of a Likert scale (1-5), 1 being incomprehensible and 5 being flawless, or a summary of their thoughts in a few words. Three out of six participants rated the engine on a Likert scale, with scores ranging from 3 to 4.5.
In 4/6 reviews, the issue of terminology was raised. The Irish eTranslation engine did not seem to choose the correct term or the term was mistranslated. For example, in one case for the technical term ‘swept-path’ the engine translated it as a path that was swept by a brush, which is too literal. There were also reports of issues related to usability of the tool, with one participant reporting that it was frustrating having to log in again after some time.
6/6 participants liked the engine and 5/6 thought it was a good translation tool and useful for getting a general gist of a text, but more fine-tuning was needed. Overall, participants agreed that the Irish eTranslation engine was a useful tool in aiding the translation process, with one participant claiming that they did not expect the engine to be as effective.
Although the Irish engine may have mistranslated one or two technical terms, paragraphs were translated well overall. There are also many advantages of using the eTranslation tool as a translation aid. Firstly, a full document can be uploaded and translated within seconds, which really speeds up the translation process, and second it is more secure than other free online machine translation engines such as Google Translate. It also must be mentioned that there is now an eTranslation plug-in for WordPress which is particularly useful for translating website content.
The NRS hopes to continue to collect Irish language data from public organisations and highlight the importance of contributing data in order to improve the Irish eTranslation engine. It is clear that more data needs to be collected from various domains, in order for the Irish eTranslation engine to recognise and understand more technical terminology. The NRS will continue to promote eTranslation through its Twitter, Instagram and LinkedIn channels and gather more feedback in the New Year.