Multilingual Word Rains: Text Visualisation Built on Cross-Language Word Similarities

Abstract

In this study, the text visualisation technique Word Rain is extended with functionality for generating multilingual word rains. Using multilingual pre-trained word embeddings provided by the MUSE library, we show that cross-language studies on lexical frequency with Word Rain are possible in the same way as has previously been possible for monolingual corpora. We make the programming code for generating multilingual word rains available, and we also provide a web page where visualisations can be produced by simply uploading text files that the user wants to visualise. To illustrate the technique, a tentative study of common word groups and crosslanguage frequency similarities and differences is performed on debates from three parliaments included in the CLARIN ParlaMint corpus.