20,499 datasets found

Natural Questions (NQ) is a dataset of Google search queries with answers from Wikipedia pages provided by human annotators.

SQuAD 2.0 is a dataset of questions from Wikipedia passages, proposed by human annotators while viewing these Wikipedia passages.

A dataset comprised of synthetic parallel sentences generated from English monolingual data through translation.

The Chinese monolingual corpus used for training models, selected based on quality metrics.

The English monolingual corpus used to create synthetic data for training models by back-translation.

The bilingual training corpus for Chinese to English translation, consisting of parallel sentences selected to maximize translation quality.

The NIST Chinese-to-English dataset includes bilingual sentence pairs for training, development, and evaluation in neural machine translation.

The IWSLT 2017 dataset is used for German to English translations, focusing on spoken language translations for machine translation tasks.

The KFTT dataset is used for the Japanese-English machine translation task, facilitating the evaluation of translation performance.

The dataset consists of sensor readings collected from a music glove instrument, including pressure sensors, flex sensors, and IMU data, alongside MIDI outputs from a connected...

The Google Distant Supervision (GDS) dataset is an extension of the Google relation extraction corpus with additional instances from entity pairs.

The dataset comes from WMT18 training data and includes randomly selected samples of human references and machine translations alongside their sources for training purposes.

The dataset consists of translations produced by a state-of-the-art neural machine translation (NMT) Transformer model. It follows the WMT14 data setup, optimized on the test...

IMST-UD is the Turkish Dependency Treebank which contains syntactically annotated data used for extracting dependencies and parsing tasks.

Google Universal Dependency Treebanks provide syntactic annotations that are used for training and evaluation across various languages including Indonesian, Korean, and Japanese.

The EuroParl dataset consists of multi-parallel sentences in several European languages, used for training and evaluation with automatic Universal Dependencies annotations.

The CoNLL 2003 dataset is used for Named Entity Recognition (NER), containing annotations for four types of named entities.

The CoNLL 2012 shared task dataset is designed for coreference resolution, consisting of articles annotated with mentions that refer to the same entities.

The SNLI dataset consists of human-written English sentences annotated with the labels entailment, contradiction, or neutral, aimed at measuring textual entailment.

The One Billion Word Benchmark is a dataset used for measuring progress in statistical language modeling, featuring a large unannotated text corpus.