Category: Research

A couple of Norman Maclean’s impromptu recordings provide the material for trying out the latest developments in Gaelic Automatic Speech Recognition.

When Norman kindly offered to record some stories for Island Voices in 2015, he was perhaps already beginning to feel his age. As a result, while he delivered them all with his trademark panache, he opted in the main to read them aloud from pre-prepared scripts. This was a blessing in disguise for Island Voices, as it meant that ready-made transcripts already existed which could be easily transferred to the Clilstore platform, enabling word-by-word clickable translations – all available on our Norman Maclean page.

However, the man was irrepressible, and once he was into his stride he just kept going, meaning a couple more stories were added to the collection off the top of his head. Lively recordings resulted, but without written transcripts – until now. Island Voices has recently been working closely with the Automatic Speech Recognition project in Edinburgh, and a good number of our Gaelic films now offer optional subtitling. These were created by feeding already existing transcripts into the text aligner tool the ASR team developed as part of their work, so that individual subtitles would appear at the right time on the videos. These were texts that had been created by someone sitting down with the recording and manually typing out every word they heard – time-consuming work!

What’s new with these recordings is that it is the ASR tool on which the Edinburgh team are working that has actually itself created the first draft of the transcripts used to produce the subtitles in these films. The results were by no means perfect, and there was still a need for a human ear and hand to tidy them up before they could be used, but it’s a developmental process. And progress is clearly being made, to the extent that the Clilstore gap in our “Sgeulachdan Thormoid” collection has now been compensated for by enabling optional onscreen subtitles on the two extra recordings he made for us – “Mìorbhailteann ann am Barraigh“, and “Bodach nan Serviettes“.

And again, as with earlier versions, once the Gaelic subtitles are in the YouTube system, automatic translation into scores of other languages via Google Translate then becomes instantly available. Norman, among his many other talents, was an enthusiastic linguist – and no slouch with a computer. True, he expressed his reservations about the development of “text talk”, but we can surely allow ourselves to think that this latest technical innovation, with the human voice at its centre (his own!), would have met with his approval.

One for the Barraich:

And the other for the Hearaich:

Calum Alasdair Fraser, from Tolastadh a’ Chaolais, talks to Maggie Smith about his family connections and upbringing on the west side of Lewis.

In his childhood he spent a lot of time outdoors. Though still young himself, he remembers the sound of weaving from many houses – a sound that is no longer heard. After leaving school he tried various jobs in different locations. Now he works on Gaelic issues with An Lanntair in Stornoway.

He talks about how he values the Gaelic language and its close connection to where he lives, even though it wasn’t his home language growing up. He also talks about his musical activities, and his interest in maintaining local interest in Gaelic, relating it to a changed way of life that still has close connections to previous generations.

A full transcript with the video embedded can be viewed on the Clilstore platform here: https://multidict.net/cs/9331

Five years ago Island Voices posted about “Google Gaelic“, in wake of the excitement over the recent addition of Scottish Gaelic to the list of languages supported by Google Translate. The move, while welcomed in some quarters, was not an uncontroversial one, with many people voicing fears over how it could be easily misused, for example to the detriment of people working in translation services.

The pitfalls inherent in using Google Translate, in particular to try to generate well-formed Gaelic from English input, are still very much there. However, here’s a good news story relating to translation from Gaelic into English. Our five-year old post gave a sample of English text generated by processing the Gaelic voiceover script for our peatcutting documentary through Google Translate. Five years later we’ve just done the same thing again.

You can compare the two texts here, with 2016 in the left column, and 2021 in the right. The later one is still by no means perfect, but it’s still clearly an improvement on the first version. Google is “learning”, and Island Voices welcomes that, particularly in light of the recent adoption of optional CC subtitles on our Series 1 and Series 2 films, meaning the original Gaelic can be machine translated into scores of other languages at the click of a mouse…

Maggie Smith, from Achmore on Lewis, has been doing a power of work collecting and recording Gaelic stories and poetry around and about Lewis for a number of years, many of them curated on her own website, and reproduced on the Island Voices page dedicated to her work. Nor has lockdown stopped her, as she reveals in this conversation with Pàdruig Moireach conducted over Zoom.

This is a new and experimental departure for the Stòras Beò nan Gàidheal team, seeking to make a virtue out of necessity. Indeed, in some ways community recording work may become easier as more and more of us get accustomed to using technology to overcome physical barriers. If this works well, we can expect more of this kind of material in the months to come.

In the first part, Maggie talks about early childhood memories and stories of Glasgow where she was born, though her Achmore roots go back many generations. Returning home she recalls the kind of upbringing island children of her age received, in which community links and mutual responsibilities were strong. Grandparental stories from work experience in Patagonia, and snatches of Spanish at the fank guarded against cultural introversion. She recalls her schooling, and the impact of television’s arrival on cèilidh culture, with traditional work on the land noticeably falling off in the 80s, particularly after oil work began.

A wordlinked transcript, with the video embedded, is available on Clistore here: https://multidict.net/cs/9169

In the second part Maggie and Pàdruig talk about trends in island work patterns over the years. The advent of the Arnish yard led to skills development opportunities for men across Lewis, which many later put to use in openings around the world. Weaving was a traditional occupation, frequently practised in combination with other jobs. Even as a schoolchild Maggie was accustomed to fitting her schoolwork into other duties, such as fetching water for the house. After a short spell working in Inverness after school, she returned to work with the family haulage firm for many years, before branching out into media work, tourism and other projects.

A wordlinked transcript, with the video embedded, is available on Clilstore: https://multidict.net/cs/9170

In the third part Maggie talks more about her cultural activity in the community, including community drama based on locally sourced stories, and the collection of local poetry. Moving to Zoom during lockdown has created a new platform for locals to share stories and for incomers and Gaelic learners to learn about the culture, recreating old communities and gathering new people. She also talks about the power of music and song in working with older people at risk of memory loss, and of collecting fishermen’s stories, mostly in Gaelic. The conversation ends with a discussion of changes that have come over Achmore and the use of Gaelic in the community.

A wordlinked transcript, with the video embedded, is available on Clilstore: https://multidict.net/cs/9171

Gàidhlig agus Beurla, gu cinnteach – ach a bheil cànain eile aig Flòraidh NicDhòmhnaill? Agus ma tha, cia mheud!?

Abair seachdain “techie” a th’ air a bhith aig Guthan nan Eilean. Bha fèill mhòr Dihaoine is Disathairne sa chaidh air na fo-thiotalan “automataigeach” a chaidh a chruthachadh aig Oilthigh Dhùn Èideann ann am pròiseact Shoillse, ach dè eile a thàinig a-mach à sin ach cothrom fo-thiotalan a chur air na bhidiothan Gàidhlig uile gu lèir a th’ aig Guthan nan Eilean air YouTube! Cha ghabh an obair sin dèanamh taobh-a-staigh latha, ach tha an sgioba ris a’ ghnothach, agus tha Sreath 1 deiseil mar-thà.

Chan e sin deireadh an sgeòil ge-tà. Le fo-thiotalan “san t-siostam” a-nis tha sin a’ fàgail gur urrainnear “eadar-theangachadh” a thabhann cuideachd tro Google Translate air na fo-thiotalan sna bhidiothan. Cha bhi iad gun mhearachd idir, ach can nam biodh càirdean agad aig nach eil Gàidhlig: an dèidh dhut “CC” a chur air, faodar an uair sin na settings air “subtitles” atharrachadh gu “auto-translate” airson tionndadh air choireigin a thabhann dhaibh ann am Beurla – no Frangais, Gearmailtis, agus iomadh cànan eile.

Seo Flòraidh, ma tha, ann an Sreath 1, ri “leughadh” cha mhòr ann an cànan sam bith a thogras tu.

Agus mar chuimhne air a’ chuspair air a bheil i a’ bruidhinn, cuir sùil a-rithis air a’ phost “Community Adaptations” airson tionndaidhean eile (gun fo-thiotalan) fhaicinn dhe na filmichean aithriseach a thòisich an còmhradh, le seann charaid eile aig Guthan nan Eilean na rionnag annta…