View Proposal

Proposer
Beatrice Alex
Title
Named Entity Recognition and Geoparsing for Scottish Gaelic Text
Goal
This project aims to develop and evaluate a named entity recognition (NER) tool for Scottish Gaelic that identifies person and location names in text, and to extend it into a geoparsing pipeline that ground and resolves location mentions to real-world coordinates. The contribution lies partly in the tool itself and partly in the evaluation to understand what methods work well for transferring this task to a low-resource and morphologically rich language.
Description
NER and geoparsing are standard information-extraction tasks: NER locates and classifies mentions of people, places and other entities, while geoparsing resolves place name mentions to geographic coordinates. Together they turn unstructured text into structured and mappable data, enabling better data access and analysis. For Scottish Gaelic, a low-resource language, such tool do not yet exist. The project will use the Annotated Reference Corpus of Scottish Gaelic (ARCOSG) [1] as its starting point. ARCOSG is a representative, hand-tagged corpus of roughly 86k tokens across eight spoken and written registers. Its part-of-speech layer marks proper nouns (differentiated into person and location names), though it contains no geographic grounding. Some additional annotation will therefore be required for the geo-resolution part of the project. The dataset is small and can be used as a seed for training / augmentation and for evaluation. Traditionally, geoparsing is made of two steps: 1) an initial NER step (detecting place-name mentions) and 2) a geo-resolution (disambiguating them and assigning coordinates, typically against a gazetteer). An example geoparser which works in this way for English text is the Edinburgh Geoparser [2]. Gaelic raises specific challenges worth considering: • initial consonant mutation alters the surface form of names (e.g. Glaschu (Glasgow) becomes Ghlaschu after certain prepositions), • many places have both Gaelic and anglicised forms, and • gazetteer coverage of Gaelic toponyms is uneven, though resources such as Ainmean-Àite na h-Alba [3], Ordnance Survey [4] and GB1900 [5] offer partial coverage. Students taking on this project need training in NLP (namely F20NL/F21NL) with experience working with textual data; an interest in linguistic annotation would be helpful. Proficiency in Gaelic is not required; domain knowledge and place-name advice can be sought from Prof Will Lamb, an expert in Scottish Gaelic at the University of Edinburgh and an active collaborator of the supervisor. Students should contact the supervisor if they are interested in pursuing research in this area, ideally with a more concrete idea of which aspect they would like to focus on (e.g. NER only, or the full geoparsing pipeline). Students may also propose their own project in the area of low-resource languages and NLP/AI, in which case they should contact the supervisor with details of their project idea and the dataset they would like to work with.
Resources
[1] Lamb, William et al. (2020). Annotated Reference Corpus of Scottish Gaelic (ARCOSG), 1997–2020 [dataset]. University of Edinburgh. https://github.com/Gaelic-Algorithmic-Research-Group/ARCOSG [2] Grover, Claire, et al. "Use of the Edinburgh geoparser for georeferencing digitised historical collections." Philosophical Transactions of the Royal Society A 368.1925 (2010): 3875–3889. [3] https://www.ainmean-aite.scot [4] https://osdatahub.os.uk [5] https://data.nls.uk/data/map-spatial-data/gb1900/
Background
Url
Difficulty Level
Moderate
Ethical Approval
Full
Number Of Students
2
Supervisor
Beatrice Alex
Keywords
natural language processing, nlp, geoparsing, scottish gaelic, evaluation
Degrees
Bachelor of Science in Computer Science
Master of Science in Artificial Intelligence
Master of Science in Artificial Intelligence with SMI
Master of Science in Data Science
Bachelor of Science in Computing Science
Back to List