Alts, Abbreviations, and AKAs: Historical Onomastic Variation and Automated Named Entity Recognition

Abstract

The accurate automated identification of named places is a major concern for scholars in the digital humanities, and especially for those engaged in research that depends upon the gazetteer-led recognition of specific aspects. The field of onomastics examines the linguistic roots and historical development of names, which have for the most part only standardised into single officially recognised forms since the late nineteenth century. Even slight spelling variations can introduce errors in geotagging techniques, and these differences in place-name spellings are thus vital considerations when seeking high rates of correct geospatial identification in historical texts. This article offers an overview of typical name-based variation that can cause issues in the accurate geotagging of any historical resource. The article argues that the careful study and documentation of these variations can assist in the development of more complete onymic records, which in turn may inform geotaggers through a cycle of variational recognition. It demonstrates how patterns in regional naming variation and development, across both specific and generic name elements, can be identified through the historical records of each known location. The article uses examples taken from a digitised corpus of writing about the English Lake District, a collection of 80 texts that date from between 1622 and 1900. Four of the more complex spelling-based problems encountered during the creation of a manual gazetteer for this corpus are examined. Specifically, the article demonstrates how and why such variation must be expected, particularly in the years preceding the standardisation of place-name spellings. It suggests how procedural developments may be undertaken to account for such georeferential issues in the Named Entity Recognition strategies employed by future projects. Similarly, the benefits of such multi-genre corpora to assist in completing onomastic records is also shown through examples of new name forms discovered for prominent sites in the Lake District. This focus is accompanied by a discussion of the influence of literary works on place-name standardisation-an aspect not typically accounted for in traditional onomastic study-to illustrate the extent to which authorial interests in regional toponymic histories can influence linguistic development.

A pragmatic guide to geoparsing evaluation

“…Twitter datasets such as GeoCorpora (Wallgrün et al 2018) experience a gradual decline in completeness as users delete their tweets and deactivate profiles. WoTR DeLozier et al 2016and CLDW Rayson et al (2017) are suitable only for digital humanities due to their historical nature and localised coverage, which is problematic to resolve (Butler et al 2017). CLUST (Lieberman and Samet 2011) is a corpus of clustered streaming news of global events, similar to LGL.…”

Hearing the silence: finding the middle ground in the spatial humanities? Extracting and comparing perceived silence and tranquillity in the English Lake District

“…In the case of Geograph, extracting sentences is a relatively simple task, since the texts largely consist of short captions for the associated image. The CLDW, on the other hand, presented more challenges, including idiosyncratic punctuation, case and hyphenation (Butler et al 2017. Nonetheless, in both cases sentence extraction and tokenisation were carried out using the NLTK Python Library with no modifications.…”

Mapping Landscapes in Transformation

“…These coordinate data are allocated through additional XML tags that are added to the texts. Conducting this process in an entirely automated manner was found not to be satisfactory for the complex place-names found in the CLDW (see Butler et al 2017). The process was enhanced using concordance geoparsing -where a small subset of the text is geoparsed, the results are checked, and any corrections fed into processing subsequent subsets (Rupp et al 2014) -and a considerable amount of manual checking.…”