This resource contains a pre-trained BERT language model trained on the Portuguese language. A BERT-Large cased variant was trained on the BrWaC (Brazilian Web as Corpus), a large Portuguese corpus, for 1,000,000 steps, using whole-word mask. The model is available as artifacts for TensorFlow an...
This resource contains model weights for five Transformer-based models: RoBERTa, GPT-2, T5, BART and COMET(BART). These models were implemented using HuggingFace, and fine-tuned on the following four commonsense reasoning tasks: Argument Reasoning Comprehension Task (ARCT), AI2 Reasoning Challen...
We publish three novel datasets called N3. N3 will be published using NIF ensuring a greater interoperability to overcome the need for corpus-specific parsers. The data can be downloaded from our project homepage.
This set of materials pertains to a study on the production of explicit pronouns, null pronouns, and repeated-NP anaphors, in European Portuguese. A spreadsheet containing data from 73 participants (young adults), namely, count data for instances of the different types of anaphor that occurred in...
PS corpus (Post-Scriptum) - treebank is a treebank corpus of 586 informal mail letters written in Portuguese and Spanish during the Modern Ages (from the XVIth century to the beginning of the XIXth century). This treebank is a syntactically annotated subset of the Portuguese "PS corpus (Post-S...
Multilingual (CEF languages) corpus acquired from the website https://antibiotic.ecdc.europa.eu/ . It contains 20981 TUs (in total) for EN-X language pairs, where X is a CEF language.
Multilingual (CEF languages) corpus acquired from website (https://ec.europa.eu/commission/presscorner/) of the EU portal (8th July 2020). It contains 23 TMX files (EN-X, where X is a CEF language) with 151895 TUs in total.
XGLUE is a new benchmark dataset to evaluate the performance of cross-lingual pre-trained models with respect to cross-lingual natural language understanding and generation. XGLUE is composed of 11 tasks spans 19 languages. For each task, the training data is only available in English. This me...
The DepBankPT (Branco et al., 2011a) is a corpus of grammatical dependencies of the translated news composed of 3,406 sentences and 44,598 tokens taken from the Wall Street Journal. The DepBankPT is aligned to a constituency bank, the TreeBankPT (see Branco et al., 2011b). The key bridging eleme...
A corpus of 2,000 MEDLINE abstracts, collected using the three MeSH terms human, blood cells and transcription factors. The corpus is available in three formats: 1) A text file containing part-of-speech (POS) annotation, based on the Penn Treebank format, 2) An XML file containing inline POS anno...