Visualising Typological Relationships: Plotting WALS with Heat Maps

Proceedings of the EACL 2012 Joint Workshop of LINGVIS & UNCLH, pages 30–34, Avignon, France, April 23 - 24 2012. c⃝2012 Association for Computational Linguistics Visualising Typological Relationships: Plotting WALS with Heat Maps Richard Littauer University of Saarland Computational Linguistics Saarbr¨ucken, Germany richard.littauer@gmail.com Rory Turnbull Ohio State University Department of Linguistics Columbus, Ohio turnbull@ling.osu.edu Alexis Palmer University of Saarland Computational Linguistics Saarbr¨ucken, Germany apalmer@coli.uni-sb.de Abstract This paper presents a novel way of vi- sualising relationships between languages. The key feature of the visualisation is that it brings geographic, phylogenetic, and linguistic data together into a single im- age, allowing a new visual perspective on linguistic typology. The data presented here is extracted from the World Atlas of Language Structures (WALS) (Dryer and Haspelmath, 2011). After pruning due to low coverage of WALS, we filter the typo- logical data by geographical proximity in order to ascertain areal typological effects. The data are displayed in heat maps which reflect the strength of similarity between languages for different linguistic features. Finally, the heat maps are annotated for lan- guage family membership. The images so produced allow a multi-faceted perspective on the data which we hope will facilitate the interpretation of results and perhaps illumi- nate new areas of research in linguistic ty- pology. 1 Introduction This paper presents a novel way of visualising re- lationships between languages. Relationships be- tween languages can be understood with respect to linguistic features of the languages, their geo- graphical proximity, and their status with respect to historical development. The visualisations pre- sented in this paper are part of a new attempt to bring together these three perspectives into a sin- gle image. One line of recent work brings com- putational methods to bear on the formation and use of large typological databases, often using so- phisticated statistical techniques to discover rela- tions between languages (Cysouw, 2011; Daum´e III and Campbell, 2007; Daum´e III, 2009, among others), and another line of work uses typolog- ical data in natural language processing (Georgi et al., 2010; Lewis and Xia, 2008, for example). The task of visually presenting the resulting data in this way has been only infrequently addressed. We are aware of some similar work (Mayer et al., 2010; Rohrdantz et al., 2010) in visualising dif- ferences in linguistic typology, phylogeny (Mul- titree, 2009), and geographical variation (Wiel- ing et al., 2011). Here, we present our method for addressing the visualisation gap, bringing to- gether phylogeny, typology, and geography by us- ing data from the World Atlas of Language Struc- tures (Dryer and Haspelmath, 2011) to develop heat maps that can visually show the intercon- nected relationships between languages and lan- guage families. The main envisioned application of our visual- isations is in the area of linguistic typology. Ty- pology has been used to derive implications about possible languages, and about the ordering of the human mind. Different theorists have taken dif- ferent views on the relationship between typology and the universality of languages. For example, Greenberg (1963), a foundational …

Content truncated to 500 words