Greek BERT

Available

• https://github.com/nlpaueb/greek-bert

Team

• NLP Group of Athens University of Economics and Business: NLP AUEB on Github

Author

• Ilias Chalkidis on Github

Description

From the project, "A Greek edition of Google's BERT pre-trained language model."[1] BERT stands for "Bidirectional Encoder Representations from Transformers", "a new method of pre-training language representations which obtains state-of-the-art results on a wide array of Natural Language Processing (NLP) tasks."[2] BERT was first released in 2018 by Google as the first unsupervised, bidirectional NLP model, with an architecture similar in certain ways to later GPT (Generative Pre-training Transformer) based models.[3] Building on BERT, Greek BERT can be used for tasks such as sentence classification, question answering,

Nota bene: this language model applies to modern Greek and is trained on modern Greek corpora (from Wikipedia, European Parliament, OSCAR, etc). It is not intended for working with Classical, Post-Classical, or Byzantine Greek corpora.