Corpus Annotation and Data Analysis (CAnDA) An Equinox School @ Gauss's Observatory

Abstract
Recent years have seen a boom in corpus-based linguistic research on sign languages, following the creation of multiple sign language corpora. These corpora are invaluable for the documentation of sign languages as well as continued linguistic research on their grammatical structure and degree of intra-linguistic variation. At the same time, given the relatively small size of all existing sign language corpora, the time-consuming annotation process, and the present scarcity of automatic annotation tools, using sign language corpora for linguistic research involves different challenges than working with (large) (majority) spoken language corpora. We will look at examples from the literature to explore what type of research questions are ideally tackled by corpus-based studies on sign languages. You will then get some hands-on experience working with data from the German Sign Language Corpus. Using ELAN Linguistic Annotator, you will create and analyze your own annotations in a small subset of the data in this corpus to answer a basic research question. Familiarity with sign languages and sign language linguistics is not required.

Requirements/Preparation
In this course, we will be using ELAN Linguistics Annotator, which you will have to install before the course starts. The latest version can be downloaded here: https://archive.mpi.nl/tla/elan. While you’re at it, feel free to peruse the full manual and/or how-to guide under the ‘documentation’ tab.
We will be working with data from the German Sign Language Corpus (DGS Corpus), available at https://www.sign-lang.uni-hamburg.de/meinedgs/ling/start_en.html. I recommend reading the general information on the homepage prior to the start of the course.

Materials
Some 50 hours of material from the DGS Corpus, including transcription files, are freely available on the DGS Corpus website. A selection of these data will be used during the course and will be made available to you in due time.

Abstract

This class will give an introduction into geolinguistics, which is a branch of linguistics and geography. We will look into social media text messages that are associated with location information. This location information can be expressed by an exact address, e.g. Humboldtallee 19, Göttingen. We will first look at language distribution in multilingual countries to trace language borders, e.g. where French is spoken in Belgium and where Dutch is spoken in the same country. We will do the same with multilingual cities such as the city of New York and see what languages are spoken (the most) in Chinatown and other parts of Manhattan, The Bronx, and Queens. We will then look at the distribution of regional language varieties or dialects and trace the border of the dialectal word Semmel ‘bread roll’ and Brötchen ‘bread roll’ in German. Finally, we will visualise the distribution of loanwords from indigenous languages in South America (e.g. pucho ‘cigarette’, cancha ‘pitch’, palta ‘avocado’, etc.). Students will learn the techniques required to visualise language and dialect distribution on geographic maps.

Materials and recommendations

In this particular course, we are going to learn how to work with geolocation associated with natural language data from social media platforms like Twitter.

I recommend to getting familiar with geolocation information on Twitter:

https://www.tweetbinder.com/blog/twitter-geolocation-map/
https://developer.twitter.com/en/developer-terms/geo-guidelines
introduction of the article in https://par.nsf.gov/servlets/purl/10106302

For application of geolocation in linguistics, I recommend reading at least Mocanu et al. 2013:

Bland Justin & Terrel A. Morgan, 2020, Geographic variation of voseo on Spanish Twitter. Guillermo Lorenzo (ed.) Issues in Hispanic and Lusophone Linguistics 27. 7-38. John Benjamins.
Gonçalves, Bruno & David Sánchez, 2014, ‘Crowdsourcing dialect characterization through Twitter’, PloS ONE 9: e112074.
Grieve, Jack; Chris Montgomery; Andrea Nini; Akira Murakami & Diansheng Guo, 2019, ‘Mapping lexical dialect variation in British English Using Twitter’, Front. Artif. Intell. 2(11). doi: 10.3389/frai.2019.00011.
Lansley G, Longley PA, 2016, ‘The geography of Twitter topics in London.’ Comput Environ Urban Syst. 58:85–96.
Leemann, Adrian; Marie-José Kolly; Ross Purves; David Britain & Elvira Glaser, 2016, ‘Crowdsourcing Language Change with Smartphone Applications’, PLoS ONE 11(1): e0143060. doi: 10.1371/journal.pone.0143060.
Levy Abitbol, Jacob; Márton Karsai; Jean-Philippe Magué; Jean-Pierre Chevrot & Eric Fleury, 2018, ‘Socioeconomic dependencies of linguistic patterns in Twitter: a multivariate analysis’, Proceedings of the 2018 World Wide Web Conference WWW’18, 1125–1134.
Mocanu, Delia; Baronchelli, Andrea; Perra, Nicola; Gonçalves, Bruno; Zhang, Qian & Vespignani, Alessandro, 2013, The Twitter of Babel: Mapping World Languages through Microblogging Platforms. PLoS ONE 8(4): e61981. DOI: 10.1371/journal.pone.0061981
Wieling M, Nerbonne J, Baayen RH, 2011, Quantitative Social Dialectology: Explaining Linguistic Variation Geographically and Socially. PLoS ONE 6(9): e23613. https://doi.org/10.1371/journal.pone.0023613

Maps and software packages we will use for geolocation analysis, I recommend studying:

OpenStreetMap. OpenStreetMap License; 2017. Available from: http://wiki.openstreetmap.org/wiki/Open_Database_License.
Met Office. 2010–2015. Cartopy: a cartographic python library with a Matplotlib interface. Online: https://scitools.org.uk/cartopy/ (last access: 24.01.2022). https://developer.twitter.com/en/docs/tutorials/filtering-tweets-by-location

In this course we will briefly overview the core dynamics of the eye tracking methodology applied to psycholinguistic investigations. By presenting some experimental works employing silent reading during eye tracking (first part) and visual world paradigm (second part) we will discuss the building blocks of an eye tracking study: how it is implemented, how the results can be analysed, what hypotheses it is geared to test, and what the main motivations behind choosing this technique are.
 

Eye tracking, Day 1: reading

Eye tracking, Day 2: visual world

This course introduces participants to the world of computational corpus linguistics. Especially, this course focuses on the cutting-edge technologies and frameworks in this field such as Universal Dependencies and BERT (or RoBERTa, T5, etc.). Participants will build and analyze corpora on their own computers, so they need to bring laptops. This course includes many hands-on sub-sessions in groups. The groups will present their results at the end of the course.
Participants will acquire knowledge in the history of computational corpus linguistics and current trends of computational corpus linguistics and related fields such as natural language processing. Students will become accustomed to using and making linguistically tagged corpora, Universal Dependencies treebanks, and natural language processing tools such as Transformer and BERT models.
 

Slides at So's homepage