The RELATE platform is an extensible web application portal comprising a set of NLP components for the Romanian language, such as sentence segmentation, tokenization, part of speech tagging, word embeddings calculation, wordnet linking, named entity recognition, etc.
The main RELATE modules are:
1. TEPROLIN is a web service exposing basic NLP processing tools, such as: text normalization; diacritics restoration; word hyphenation; phonetic transcription using the SAMPA phonemes for Romanian; numeral and abbreviation rewriting; sentence splitting; tokenization; POS tagging; lemmatization; named entity recognition; biomedical NER; chunking; dependency parsing.
2. The Reference Corpus for Contemporary Romanian Language (CoRoLa) constructed between 2014 and 2017. It contains both written texts and oral recordings. Its aim was to cover major functional language styles (legal, scientific, journalistic, imaginative, memoirs, administrative), in four domains (arts and culture, nature, society, science).
3. The Romanian wordnet RoWordNet (59348 synsets covering 85277 literals)
4. The Speech Synthesis for Lightweight Applications (SSLA), available as a web endpoint, allowing the production and download of a wav file with the synthesized text.
