We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers.
SOTA for Common Sense Reasoning on SWAG
We make all code and pre-trained models available to the research community for use and reproduction.
#4 best model for Named Entity Recognition on CoNLL 2003 (English) (using extra training data)
We present FLAIR, an NLP framework designed to facilitate training and distribution of state-of-the-art sequence labeling, text classification and language models.
fairseq is an open-source sequence modeling toolkit that allows researchers and developers to train custom models for translation, summarization, language modeling, and other text generation tasks.
Pre-trained word vectors are ubiquitous in Natural Language Processing applications.
Generalization and reliability of multilingual translation often highly depend on the amount of available parallel data for each language pair of interest.
In this paper, we describe compare-mt, a tool for holistic analysis and comparison of the results of systems for language generation tasks such as machine translation.
Aspect-based sentiment analysis (ABSA), which aims to identify fine-grained opinion polarity towards a specific aspect, is a challenging subtask of sentiment analysis (SA).