Workflow Reversal and Data Wrangling in Multilingual Diachronic Analysis and Linguistic Linked Open Data Modelling

Abstract

The article deals with data wrangling in a multilingual collection intended for diachronic analysis and linguistic linked open data mod- elling for tracing concept change over time. Two types of static word embeddings are used: word2vec (French and Hebrew data sets), and fastText (Latin and Lithuanian data sets). We model examples from these embeddings via the OntoLex-FrAC formalism. To address the challenge of heterogeneity, we use a minimalist workflow design allowing for both convergence and flexibility in attaining the project goals.