incom.py 2.0 - Calculating Linguistic Distances and Asymmetries in Auditory Perception of Closely Related Languages

Proceedings of Recent Advances in Natural Language Processing, pages 968–977 Sep 1–3, 2021. https://doi.org/10.26615/978-954-452-072-4_110 968 incom.py 2.0 – Calculating Linguistic Distances and Asymmetries in Auditory Perception of Closely Related Languages Marius Mosbach1,2, Irina Stenger1, Tania Avgustinova1, Bernd Möbius1, Dietrich Klakow1,2 1 Department of Language Science and Technology, Saarland University, Germany 2 Saarland Informatics Campus {mmosbach, dlakow}@lsv.uni-saarland.de ira.stenger@mx.uni-saarland.de {avgustinova, moebius}@lst.uni-saarland.de Abstract We present an extended version of a tool developed for calculating linguistic dis- tances and asymmetries in auditory per- ception of closely related languages. Along with evaluating the metrics available in the initial version of the tool, we introduce word adaptation entropy as an additional metric of linguistic asymmetry. Potential predictors of speech intelligibility are vali- dated with human performance in spoken cognate recognition experiments for Bul- garian and Russian. Special attention is paid to the possibly different contributions of vowels and consonants in oral intercom- prehension. Using incom.py 2.0 it is pos- sible to calculate, visualize, and validate three measurement methods of linguistic distances and asymmetries as well as car- rying out regression analyses in speech in- telligibility between related languages. 1 Introduction 1.1 Background Individual (receptive) multilingualism is essen- tial in the European cultural and economic area.1 How can native speakers of L1 sponta- neously understand Lx on the basis of L1-Lx relatedness? Common experience shows that the degree of success in intercomprehension differs between spoken and written modalities, due to various linguistic and non-linguistic fac- tors (Gooskens, 2019). While in listening the time available for the auditory input processing is limited, in reading one can jump back at will (Möller and Zeevaert, 2015) during visual input processing. The latter scenario, however, ex- cludes the possibility to check comprehension 1(https://ec.europa.eu/education/policies/ multilingualism/about-multilingualism-policy_ en) through interactive communicative feedback. Even though cross-lingual differences are a mul- tidimensional phenomenon (van Heuven, 2008), a simple baseline prediction of speech intelligi- bility between related languages can be based on the similarity of their phoneme inventories and lexicons. For example, both similarity of phoneme inventories and lexical similarity were identi- fied as factors predicting the perceptual confus- ability (and, by implication, similarity) of lan- guages in the Great Language Game (Skirgård et al., 2017). These authors report confusion asymmetries within the set of Slavic languages, i.e., cases of speakers of one language (e.g. Bul- garian) understanding another language (e.g. Russian) better than the other way round. Pho- netic and lexical distances are also considered as determinants of mutual intelligibility in the work of Gooskens and colleagues, who use a game-like interface, MICReLa2, to collect data for Germanic, Romance, and (six) Slavic lan- guages (Gooskens and van Heuven, 2020). Notions of entropy and surprisal (Shannon, 1948) are employed in the INCOMSLAV frame- work3 to measure information density and gauge how language users master high degrees of surprisal due to partial incomprehensibil- ity of linguistic encodings. The key idea is that understanding an unknown but related Lx should be better when the L1 language model adapted for processing the unknown Lx exhibits relatively low average surprisal, or …

Content truncated to 500 words