A Rule-based Shallow-transfer Machine Translation System for Scots and English

A Rule-based Shallow-transfer Machine Translation System for Scots and English Gavin Abercrombie jst662@alumni.ku.dk Abstract An open-source rule-based machine translation system is developed for Scots, a low-resourced minor language closely related to English and spoken in Scotland and Ireland. By concentrating on translation for assimilation (gist comprehension) from Scots to English, it is proposed that the development of dictionaries designed to be used within the Apertium platform will be sufficient to produce translations that improve non-Scots speakers understanding of the language. Mono- and bilingual Scots dictionaries are constructed using lexical items gathered from a variety of resources across several domains. Although the primary goal of this project is translation for gisting, the system is evaluated for both assimilation and dissemination (publication-ready translations). A variety of evaluation methods are used, including a cloze test undertaken by human volunteers. While evaluation results are comparable to, and in some cases superior to, those of other language pairs within the Apertium platform, room for improvement is identified in several areas of the system. Keywords: Scots, machine translation, assimilation 1. Introduction The Scots language is spoken by over 1.5 million people in Scotland1 and a further 140,000 in Ireland,2 but is low- resourced in terms of technology - as far as I am aware there are currently no automatic translators available for this lan- guage. The development of machine translation (MT) sys- tems can lead to positive outcomes for a minor language and its speakers by contributing to the language’s standard- isation, and assisting the diffusion of content originally cre- ated in that language (Forcada, 2006). The aim of this project is to take a step towards the ‘de-minorizing’ of Scots by developing a shallow transfer, rule-based MT system as part of the Apertium platform. Since all Scots speakers comprehend standard English,3 this study will focus on translation from Scots to English for assimilation (or gisting) purposes. With assimilation thought to be the most common use of MT systems today (O’Regan et al., 2013), this would seem to be an effective use of time and resources as an initial goal in MT between these two languages. As the two languages are very closely related, it is proposed that translation of the bulk of lexical items in Scots texts (combined with part-of-speech (POS) analysis) will be sufficent to enable gist comprehension by non-Scots speakers. 2. Background 2.1. The Scots language Scots is a West Germanic language descended from Old Northumbrian (a branch of Old English) and closely related to modern English (Macafee and Aitken, 2002). Indeed, it shares many characteristics with its neighbour and relative including SVO sentence structure and a ‘vast body’ of com- mon vocabulary (Tulloch, 1997). Grammatical differences can however be seen in the word order of some constructions e.g. whit fur did he dae that? 1Scotland’s Census 2011 - National Records of Scotland 22001 Census of Northern Ireland - Northern Ireland Statistics & Research Agency 3Scotland’s Census 2001 - National Records of Scotland (‘what did he do that for’). There are also notable …

Content truncated to 500 words