Within the context of the FastText initiative, pre-trained word embeddings have been made available for 294 language, based on the respective Wikipedia corpus for the particular languages. One of these languages is Aramaic, which is currently conceived of as endangered, since it is only spoken by a few minority groups in the Middle East. Nevertheless, this language has a rich history of culture and literature, being among others also the language of Jesus and the main language of the Syriac culture, which lives on today as a liturgical language in several church denominations in among others India, Syria and Iraq. This paper wants to provide first insights into the usefulness of these word embeddings to connect the separate parts of Aramaic culture, and to study them as one language with many facets and influence, a subject which hitherto has only seen separated scholarship along the lines of research questions limited to a specific time frame. Using some of the specific assets of the FastText algorithm, we show how traditional difficulties in bringing together the Aramaic literature from a computational perspective, such as limited training resources and significant lexical richness due to external influences throughout the centuries, can now accounted for.
@InProceedings{COECKELBERGS18.12, author = {Mathias Coeckelbergs}, title = {Classifying and Searching Resource-Poor Languages More Efficiently. Using the FastText Word Embeddings for the Aramaic Language Family.}, booktitle = {Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018)}, year = {2018}, month = {may}, date = {7-12}, location = {Miyazaki, Japan}, editor = {Claudia Soria and Laurent Besacier and Laurette Pretorius}, publisher = {European Language Resources Association (ELRA)}, address = {Paris, France}, isbn = {979-10-95546-22-1}, language = {english} }
Summary of the paper