Translating Text Into Irish: Using AI and Other Tools
This article is in partnership with Day Translations.
If you ever come across a social media post that is written in Irish (Gaelic or Gaeilge), you may have noticed that there is often no automatic translation available into English (or your local language). For example, if you are scrolling through Instagram or X and click through a post that is written in Irish (“in Irish” is translated into: as Gaeilge), you will not see the option of getting a translation in English or another language.
Irish Was Added to Google Translate in 2009
As you can see from this article from the Irish Times, Google added the Irish language to its Translate service back in 2009 but if you think about posts that you may have come across on Instagram, for example, that are entirely in Irish, you don’t have the option of getting them translated into other languages. This is quite strange as social media platforms such as Facebook and what was formally called Twitter (X now) were around at that time, although having automated translations came later (Twitter introduced them in 2020 and Instagram had already done so in 2016).
Back in 2009, Google Translate only supported 51 languages, so the fact that Irish was also included was a significant step, especially considering that Irish had only been a recognised EU language since 2007.
Google Translate did ask translators at the time to provide input to improve translations to and from Irish, as the initial results were less than perfect. There were some teething problems in the translations with issues including errors and omissions.
By 2019 Google Translate supported 100 languages and had introduced the image lens function for the Irish language using Neural Machine Translation technology, whereby you could translate text that was embedded in a picture, sign, or other visual form, and get it translated from Irish into your chosen language (source: Irish Independent). This was a particularly useful development, because if you are a tourist in Ireland and you happen to come across a road sign that is only in Irish without an English translation, you can immediately understand the meaning of the text.
Machine Translation for the Irish Language Is More Sophisticated Now
As of 2024, there has been tremendous progress in machine translation for the Irish language. RTÉ, the national Irish broadcaster, reported that new tools converting data to text have been developed in recent years.
In contrast with tools such as Chat GPT, which are resource-intensive, researchers have developed open source tools using their linguistic knowledge and embedding Irish grammar rules into the system. Irish can be a difficult language to learn so automation can help learners but also any other users of the language.
Irish grammar rules have similar patterns to Arabic, Spanish, Italian, Hebrew and French, but also English and German. There are conjugations to be memorised and more, especially when you need to choose a verb based on the action that is being taken, who is taking it and where. There are also variations to indicate the passing of time or whether something is permanent or temporary.
In a nutshell, you need to do a lot of studying if you want to become knowledgeable in the Gaelic language.
Gaelic was first spoken about 2,500 years ago and it has been in written form since the 5th or 6th century AD. It is one of the oldest written languages using the Roman alphabet in Northern Europe. Although Gaelic introduced some words from other languages, it remained truthful to its origins as a spoken language. Because over time English became the main language for formal written communications, in 1876 an organisation was set up to ensure the Irish language could be further studied: the Society for the Preservation of the Irish Language. In 1958 the Irish government published an official document called the Official Standard to formalise the grammar rules of the written Irish language and harmonise them with the spoken language.
Again, this is where automation can support language learners: mutations, exceptions and rules of the Irish language can be a minefield. Of course, the rigid application of rules by a machine can produce incorrect results, because context, idiomatic sentences and colloquialism can completely change the meaning of some text. Irish poses many challenges for automated translation tools, therefore having a large database of different types of sentences and examples can mitigate the error rate of translated text.
Ultimately, the role of a post-machine translation editor/reviewer is essential to iron out the kinks in a basic translation and add the right “flavour” to it.
