The Lexometer: A Shiny Application for Exploratory Analysis and Visualization of Corpus Data

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 6370–6376 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 6370 The Lexometer: A Shiny Application for Exploratory Analysis and Visualization of Corpus Data Oufan Hai, Matthew Sundberg, Katherine Trice, Rebecca Friedman, Scott Grimm University of Rochester P.O. Box 270096 Rochester, NY 14627, USA scott.grimm@rochester.edu {ohai, msundbe2, ktrice, rfried11}@u.rochester.edu Abstract Often performing even simple data science tasks with corpus data requires significant expertise in data science and program- ming languages like R and Python. With the aim of making quantitative research more accessible for researchers in the language sciences, we present the Lexometer, a Shiny application that integrates numerous data analysis and visualization functions into an easy-to-use graphical user interface. Some functions of the Lexometer are: filtering large databases to generate subsets of the data and variables of interest, providing a range of graphing techniques for both single and multiple variable analysis, and providing the data in a table format which can further be filtered as well as provide methods for cleaning the data. The Lexometer aims to be useful to language researchers with differing levels of programming expertise and to aid in broadening the inclusion of corpus-based empirical evidence in the language sciences. Keywords: R (programming language), Shiny, Data Analysis, Data Visualization, Lexical Analysis, COCA 1. Introduction Quantitative methods and corpus research have increas- ingly been used to underpin research in theoretical syn- tax as well as lexical and formal semantics. At the heart of many theoretical proposals about nominal, ver- bal or adjectival structure are particular distributional properties, which are argued to correlate with different syntactic or semantic properties. Well-known exam- ples in semantics include the count/mass distinction in nominal structure and the availability of different de- terminers (Jespersen, 1913), adjectival scalar structure and the corresponding distribution of different adver- bial modifiers (Kennedy and McNally, 2005), and ver- bal aspectual structure and the corresponding tempo- ral/adverbial modifiers (Vendler, 1967). Parallel issues arise for investigating syntactic phenomena and pro- posals for their syntactic structure, such as the differ- ent types of nominalizations and the range of distri- butional characteristics assumed to correlate with them (Grimshaw, 1990), or the now classic corpus study of the dative alternation in (Bresnan et al., 2007). The use of corpus data has determined important prop- erties of classes of nouns, verbs and adjectives and helped (in)validate a range of claims, making it of great importance to the field. At the same time, there are few tools that do not require significant expertise in a pro- gramming language such as Python or R, as well as additional familiarity with a range of code libraries and packages, such as ggplot and Plotly to aid data visu- alization. This creates a significant obstacle for mak- ing more sophisticated corpus methodologies available to those without substantial programming background, such as undergraduate and even graduate RAs, while at the same time prohibits researchers who are engaged in theoretical syntax or formal and/or …

Content truncated to 500 words