Instant Annotations – Applying NLP Methods to the Annotation of Spoken Language Documentation Corpora

The 3rd International Workshop for Computational Linguistics of Uralic Languages, pages 25–36, St. Petersburg, Russia, 23–24 January 2017. c⃝2017 Association for Computational Linguistics Instant Annotations – Applying NLP Methods to the Annotation of Spoken Language Documentation Corpora Ciprian Gerstenberger* University of Tromsø – The Arctic University of Norway Giellatekno – Saami Language Technology ciprian.gerstenberger@uit.no Niko Partanen University of Hamburg / University of Freiburg Department of Uralic Studies / Department of Scandinavian Studies niko.partanen@uni-hamburg.de Michael Rießler† ENS & PSL Research University / University of Freiburg LaTTiCe / Department of Scandinavian Studies michael.riessler@skandinavistik.uni-freiburg.de Joshua Wilbur University of Freiburg Department of Scandinavian Studies joshua.wilbur@skandinavistik.uni-freiburg.de This work is licensed under a Creative Commons Attribution–NoDerivatives 4.0 International Licence. Licence details: http://creativecommons.org/licenses/by-nd/4.0/ *The order of the names of the authors is alphabetical. †Michael Rießler’s contribution to this work has received support of TransferS (laboratoire d’excellence, program “Investissements d’avenir” ANR-10-IDEX-0001-02 PSL* and ANR-10-LABX-0099). 1 25 Abstract The paper describes work-in-progress by the Pite Saami, Kola Saami and Izhva Komi language documentation projects, all of which use similar data and techni- cal frameworks and are carried out in Freiburg and in collaboration with Ham- burg, Syktyvkar, Tromsø and Uppsala. Our projects work in the endangered lan- guage documentation framework and record new spoken language data, digitize available recordings and annotate these multimedia data in order to provide com- prehensive language corpora as databases for future research on and for endan- gered and under-described Uralic speech communities. Applying NLP methods in language documentation – specifically rule-based morphological and syntactic analyzers – helps us to create more systematically annotated corpora, rather than eclectic data collections. We propose a step-by-step approach to reach higher- level annotations by using and improving truly computational methods. Ulti- mately, the spoken corpora created by our projects will be useful for scientifically significant quantitative investigations on these languages in the future. 1 Introduction Endangered language documentation (aka documentary linguistics) aims at the pro- vision of long-lasting, comprehensive, multifaceted and multipurpose records of lin- guistic practices characteristic of a given speech community [1, 2, 3, 4]. The field has made huge technological progress in regard to collaborative tools and user in- terfaces for transcribing, searching, and archiving multimedia recordings. However, paradoxically, the field has only rarely considered applying NLP methods to more efficiently annotate qualitatively and quantitatively significant corpora. This is de- spite the fact that the relevant computational methods and tools are well-known from corpus-driven linguistic research on larger written languages and are even applied to spoken varieties of these languages. Although relatively small endangered languages are increasingly in the focus of computational linguistic research (see especially Giellatekno for Northern Saami [5] and other languages, a different approach is [6]), these projects work predominantly with written language varieties. Current computational linguistic projects on endan- gered languages seem to have simply copied their approach from already established research on the major languages, including the focus on written language. The result- ing corpora are impressively large for these minority languages and include higher- level morphosyntactic annotations. However, they …

Content truncated to 500 words