20,499 datasets found

The AQUAINT dataset is used for evaluating named entity disambiguation performance.

The MSNBC dataset is used for evaluating named entity disambiguation performance.

The AIDA-CoNLL dataset consists of annotated entities in a large corpus for named entity disambiguation tasks.

The French Question Answering Dataset (FQuAD) is a native Reading Comprehension dataset comprising questions and answers extracted from Wikipedia articles. It aims to provide a...

The English Web Treebank is part of the Universal Dependencies framework and serves as a syntactically and semantically annotated corpus for training and evaluating dependency...

WIKIHOP is a dataset constructed to require multi-hop reasoning over multiple Wikipedia paragraphs while answering entity-relation questions.

CWQ (Complex Web Questions) is a dataset involving complex web-based questions requiring multiple steps to answer.

CQ (Complex Questions) consists of complex queries from Google and is designed for answering questions from a knowledge base.

SEARCHQA is a dataset designed for reading comprehension, containing trivia questions and web snippets retrieved through Google.

CMU-MOSEI is a dataset for multimodal sentiment analysis with sentiment annotations at the sentence level, featuring a blend of audio-visual and textual data.

CMU-MOSI (CMU Multimodal Opinion Sentiment Intensity) is a dataset of multimodal language focused on multimodal sentiment analysis, containing 2199 video segments from 93...

The EmoContext task dataset consists of conversations extracted from social media, intended for emotion detection, annotated with four main emotions: Happy, Sad, Angry, and...

ClueWeb includes documents from ClueWeb09-B and queries from the TREC Web Track ad hoc retrieval task 2009-2012. The dataset consists of 200 queries with relevance judgements...

The LINNAEUS dataset is a system for species name identification in biomedical literature.

The Species-800 corpus is used for species name recognition in text.

The JNLPBA corpus serves as a benchmark forbio-entity recognition tasks.

The BioCreative V CDR task corpus is a resource for chemical disease relation extraction.

Monolingual English datasets consisting of backtranslated and parallel data used for training the translation models between English, Finnish, and Estonian.

A bilingual corpus created by triangulating English–Finnish and English–Estonian parallel data, resulting in a set of 679,252 sentence pairs used to extract cognates and improve...

The WMT 2014 English-German translation dataset consists of parallel sentences in English and German used to evaluate machine translation models.