Automated Parsing of Interlinear Glossed Text from Page Images of Grammatical Descriptions

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 2878–2883 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 2878 Automated Parsing of Interlinear Glossed Text From Page Images of Grammatical Descriptions Erich R. Round∗‡§, Jayden L. Macklin-Cordes∗, T. Mark Ellison†, Sacha Beniamine§ ∗ School of Languages and Cultures, University of Queensland, Brisbane, Australia, ‡ Surrey Morphology Group, University of Surrey, Guildford, UK, § Max Planck Institute for the Science of Human History, Jena, Germany, † Cologne Center of Language Sciences, Universit¨at zu K¨oln, Germany, e.round@uq.edu.au, j.macklincordes@uq.edu.au, tmarkellison@gmail.com, beniamine@shh.mpg.de Abstract Linguists seek insight from all human languages, however accessing information from most of the full store of extant global linguistic descriptions is not easy. One of the most common kinds of information that linguists have documented is vernacular sentences, as recorded in descriptive grammars. Typically these sentences are formatted as interlinear glossed text (IGT). Most descriptive grammars, however, exist only as hardcopy or scanned pdf documents. Consequently, parsing IGTs in scanned grammars is a priority, in order to significantly increase the volume of documented linguistic information that is readily accessible. Here we demonstrate fundamental viability for a technology that can assist in making a large number of linguistic data sources machine readable: the automated identification and parsing of interlinear glossed text from scanned page images. For example, we attain high median precision and recall (>0.95) in the identification of example sentences in IGT format. Our results will be of interest to those who are keen to see more of the existing documentation of human language, especially for less-resourced and endangered languages, become more readily accessible. Keywords: ”Information Extraction, Information Retrieval”, Less-Resourced/Endangered Languages, Morphology, Typological Databases 1. Introduction Linguistic typology is the subfield of linguistics which studies the design features of human language and the dis- tribution of such features across the languages of the world. As in other disciplines, there has been a shift in linguistic typology towards the use of algorithmic techniques applied to computational data sets, which expands the scope of questions that can be asked (Bickel, 2007). However, there is a bottleneck at the stage of data acquisition. The primary sources of data for typologists are grammatical descriptions of languages. As most of these are in hardcopy, construct- ing datasets for computational use is a lengthy, manual process. Usually, the knowledge being formalised is con- fined to questions particular to the study at hand (Cooper, 2014, p. 91), so this process of manual database construc- tion must be repeated with each new research question. A research pipeline automating the process of searching the published grammatical descriptions of the world’s lan- guages would greatly reduce the time-cost of novel dataset creation. However, such a goal requires initially a machine- readable compendium of grammars. Although the digi- tisation of linguistic resources continues apace, machine- readable grammars are available for only a small minor- ity of the world’s languages (Szymanski, 2012; Szymanski, 2013). Computational linguistics can also benefit from more lin- guistic …

Content truncated to 500 words