CTAP for Italian: Integrating Components for the Analysis of Italian into a Multilingual Linguistic Complexity Analysis Tool

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 7123–7131 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 7123 CTAP for Italian: Integrating Components for the Analysis of Italian into a Multilingual Linguistic Complexity Analysis Tool Nadezda Okinina, Jennifer-Carmen Frey, Zarah Weiss Institute for Applied Linguistics, Eurac Research, Viale Druso 1, 39100 Bolzano, Italy Department of Linguistics, University of Tübingen, Wilhelmstrasse 19, 72074 Tübingen, Germany {Nadezda.Okinina, JenniferCarmen.Frey}@eurac.edu, zweiss@sfs.uni-tuebingen.de Abstract Linguistic complexity research being a very actively developing field, an increasing number of text analysis tools are created that use natural language processing techniques for the automatic extraction of quantifiable measures of linguistic complexity. While most tools are designed to analyse only one language, the CTAP open source linguistic complexity measurement tool is capable of processing multiple languages, making cross-lingual comparisons possible. Although it was originally developed for English, the architecture has been extended to support multi-lingual analyses. Here we present the Italian component of CTAP, describe its implementation and compare it to the existing linguistic complexity tools for Italian. Offering general text length statistics and features for lexical, syntactic, and morpho-syntactic complexity (including measures of lexical frequency, lexical diversity, lexical and syntactical variation, part-of- speech density), CTAP is currently the most comprehensive linguistic complexity measurement tool for Italian and the only one allowing the comparison of Italian texts to multiple other languages within one tool. Keywords: linguistic complexity, readability, Italian, text analysis, cross-lingual analysis 1. Introduction Linguistic complexity is a core construct in Second Lan- guage Acquisition (SLA) research, where Complexity, Ac- curacy, and Fluency, also known as the CAF triad, are often used to characterize language performance (Housen and Kuiken, 2009). Over the last decade, a broad variety of complexity measures has been proposed to characterize language proficiency and its development, text readability, and writing quality (Vajjala and Meurers, 2012; Bulté and Housen, 2014; Crossley and McNamara, 2014; De Clercq and Housen, 2019). Many of these linguistic complexity measures are applicable across various languages. At the same time, advances in Natural Language Processing (NLP) allow to automatically extract the measures for a va- riety of languages. In the face of these advances, a series of automatic complexity analysis approaches have been pre- sented. There are approaches for English (McNamara and Graesser, 2012; Chen and Meurers, 2016), German (Weiss and Meurers, 2019a), French (Francois and Miltsakaki, 2012), Swedish (Pilan et al., 2016) and Portuguese (Aluisio et al., 2010) containing similar - yet not completely over- lapping - sets of complexity measures. With the creation of quantifiable operationalizations for as- pects of linguistic complexity (e.g. lexical diversity) that can be calculated automatically for various languages, cross-lingual analyses can be envisioned. However, anal- yses using the same measures for texts in various languages are still rare. While the comparability of the measures used needs to be ensured from a theoretical perspective, often the extraction of linguistic complexity indices for various languages is also technically limited. Although various lan- guage-dependent tools exist that extract …

Content truncated to 500 words