NLP for Greek language

This repo is an aggregation of sources for Greek language to tackle varios Natural Language Processing/Understanding/Generation needs.

Contents

• The language and countries
• Greek Tree Bank
• Pipeline Components
  • Accentuation and diacritics
  • Lemmatization
  • Tokenization
    • NLTK
    • Spacy
  • Other
• NLP tasks
  • Named Entity Recognition
  • Translation
  • Question Answering
• Transformers model
• Other
  • Proper nouns
  • Ancient Greek

The language and countries

Greek language is spoken by majority of population in two countries.

X	Country - ISO Language code
CY
GR

Greek Tree Bank

Morphological and syntatic annotations of Greek corpus. This Greek UD source used by many other pretrained open-source components.

Manually annotated: lemmas, dependencies, POS, features.

Genres: news, wiki, spoken

Souces: public domain, wikinews articles, European Parlament sessions texts.

Corpus size: 2521 sentences/ 61.673 tokens.

https://universaldependencies.org/treebanks/el_gdt/index.html

Pipeline Components

Accentuation and diacritics

Greek text requires accents and diacritics removal. Some new Tokenizers include this step but earliest editions doesn not. https://legacy.cltk.org/en/latest/greek.html

Lemmatization

Spacy lemmatizer (trainable lemmatizer)

JohnSnowLabs Greek lemmatizer

CLTK Greek lemmatiter

Tokenization

Depends on a sutiation we might need different corpus tokenization. Sources below include general tokenizers for word, sentence, paragraph tokenization.

NLTK

NLTK tokenizer module

Spacy

Spacy Tokenizer Also available a pipeline component for Greek language senter for Sentence segmentation.

Other

Spacy offers other helpful components: morphologizer, dependency parser, attribute ruler.

NLP tasks

Named Entity Recognition

Source	Supported labels	Link
Spacy	EVENT, GPE, LOC, ORG, PERSON, PRODUCT	Spacy models
Spark NLP
Stanza
AUEB	LOC, ORG, PERSON,	gr-nlp-toolkit transformer-based

Translation

Package	Details	Link
Spark NLP	Multilingual (wrapped from Hugging Face)
Transformers	Multilingual

Question Answering

Cross-lingual QA dataset: XQuAD

Transformers model

BERT model pretrained on Greek corpus only.

bert-base-greek-uncased-v1

Greek BERT

Other

Proper nouns

List of 144,000 Classical Greek proper nouns

Ancient Greek

Some handy stuff for Ancient Greek