GAL - Generating asemic languages using deep learning
The project is dedicated to the generation of meaningless languages. New words are to be derived from a source language. These words do not occur in the root vocabulary and therefore have no semantic content. In its audiovisual form, however, the generated language should resemble its source language as closely as possible.
For this purpose, the phonemic/graphemic rules of the language are analyzed by various machine learning algorithms of Natural Language Processing (NLP). After training with textual data has been completed, the algorithms can generate individual words and even entire texts. In order to measure the objective and subjective similarity of the generated languages with the source language, linguistic measures are applied and an empirical study is conducted.
Further information and latest news on the progress of the project can be found in the Project Blog of the university competition.
