Briggin da Gap: oorNews Scrievin Formal Scots for tae Train AI

In a recent study cryed “The Sociolinguistic Foundations of Language Modeling,” researchers Jack Grieve an his colleagues delve intae da crucial role o linguistic diversity in artificial intelligence. Published in Frontiers in Artificial Intelligence, da airticle unnerscores hoo language models reflect da varieties o language dey are trained on. Da authors argue dat da societal value an performance o dese models hing on da quality an diversity o der trainin data.

Ane o da pressin issues pynted oot in da study is da representation o da Scots leid ithin AI systems. Scots, aften wrangly viewed as a dialect o English, requires carefu haendlin tae prevent AI ootpits frae soondin informal or oot o place. Sic misrepresentation can be baith discriminatory an offensive tae Scots speakers, reinforcin hairmfu stereotypes o laa status an laa education.

In response tae dese challenges, oorNews is makin significant strides by providin a muckle corpus o formal register Scots. Dis initiative aims tae ensure dat AI ootputs in Scots are nae ainly accurate but respectfu o da language’s status an uise in formal prose. By curatin content dat reflects formal Scots, oorNews helps prevent AI frae lockin intae damagin stereotypes associated wi informal language uise.

Da integration o formal Scots in AI trainin data is essential for tae create language models dat honour an accurately represent Scotland’s linguistic diversity. Dis approach aligns wi da recommendations frae Grieve et al., advocatin for trainin corpora dat genuinely reflect da specific varieties o language bein modelled. Aa Scots spikkers are walcome tae contribute tae oorNews by suggestin ony chynges dey waant tae see in an airticle via da comments section aneath.