CoNLL-2003 Named Entity
Dataset
The most classic benchmark dataset in the field of named entity recognition, containing four types of entity annotations from Reuters news corpus, and still the gold standard for NER model evaluation.
Dataset Highlights
The most influential sequence labeling benchmark in the NLP field, cited by thousands of papers
Gold Standard Benchmark
Since its release in 2003, CoNLL-2003 has been the standard evaluation benchmark for named entity recognition tasks, widely adopted by academia and industry.
Real News Corpus
The English data comes from Reuters news reports, and the German data comes from the Frankfurter Rundschau newspaper, covering diverse texts from the real world.
Four Types of Entity Annotation
Includes four types of entities: PER (person names), LOC (location names), ORG (organization names), and MISC (miscellaneous), covering the main categories of named entities.
Bilingual Dataset
Provides two language versions in English (~22,137 sentences) and German (~12,705 sentences), supporting cross-language NER research and multilingual model evaluation.
Rich Annotation Hierarchy
Each token is annotated with part-of-speech (POS) tags, syntactic chunk tags, and named entity (NER) tags, supporting multi-task learning.
Standardized Evaluation
Provides a unified division of training, validation, and test sets, along with standardized F1 evaluation metrics, ensuring fair comparison of different models.
Applicable Scenarios
From academic research to industrial applications, NER technology is everywhere
Named Entity Recognition
Train and evaluate NER models, such as BiLSTM-CRF, BERT, RoBERTa, and other sequence labeling models
Sequence Labeling Research
As a general benchmark for sequence labeling tasks, test the labeling performance of different architectures like CRF and Transformer
Information Extraction
Extract structured information such as names, locations, and organizations from unstructured text to build knowledge graphs
Token Classification
Train Token Classification models to provide entity recognition capabilities for downstream tasks such as question answering systems and text summarization
Data Preview
The following is an example of the annotation format of the CoNLL-2003 dataset, with each column representing: word, part of speech, chunk, entity label
# word POS chunk NER EU NNP B-NP B-ORG rejects VBZ B-VP O German JJ B-NP B-MISC call NN I-NP O to TO B-VP O boycott VB I-VP O British JJ B-NP B-MISC lamb NN I-NP O . . O O Peter NNP B-NP B-PER Blackburn NNP I-NP I-PER BRUSSELS NNP B-NP B-LOC 1996-08-22 CD I-NP O The DT B-NP O European NNP I-NP B-ORG Commission NNP I-NP I-ORG said VBD B-VP O on IN B-PP O Thursday NNP B-NP O it PRP B-NP O disagreed VBD B-VP O with IN B-PP O a DT B-NP O German JJ I-NP B-MISC advisory JJ I-NP O committee NN I-NP O .
3 Steps to Get Started Quickly
Start your NER research project in just a few minutes, from browsing to modeling
Browse Datasets
View dataset details on the Ace Data Cloud platform, understand entity type definitions, annotation specifications, and metadata such as dataset splits.
Download Data
Obtain training, validation, and test set files in English and German, with CoNLL column format ready to use, no additional preprocessing required.
Load and Train
Use datasets.load_dataset("conll2003") or directly parse column format files to start training your NER model.
Start Exploring the CoNLL-2003 Dataset
The gold standard benchmark in the NER field, open research license, available immediately. Whether you are a beginner in NLP or an experienced researcher, this dataset is an essential evaluation tool.