Use Ardis AI to
label a dataset in 10 minutes

Use our NLP data,
make up the label set as you go,
and watch our AI adjust to your choices.

Start Labeling

Generate a labeled dataset that's ready for model training

Our Named Entity Recognition (NER) corpus includes data from Wikipedia's Vital Articles, the American Local News Corpus, and the CNN/DailyNews corpus.

We use the labels that you apply to recommend labels for the rest of the entities.

Once you're happy with our recommendations, download the dataset in BILUO format, ready to be used with spaCy and other NLP tools.

A screenshot of the labeling dashboard depicting a list of user-chosen labels ('person', 'organization', 'location', 'product, 'event', 'date', 'norp', 'facility', 'topic', 'religion', and 'law'), followed by a card displaying detailed information about a single, selected entity. The entity detail card includes the entity name ('LBCC'), the label recommendation ('organization'), and the entity's source sentence ('Since landing on the LBCC board in 2000, he said he has been proud of the bond initiatives passed to improve LBCC infrastructure and the fact the school recently has maintained a balanced budget despite massive cuts.').  Below the entity detail card, there is a grid containing the names of a selection of entities from the data set. The grid element of the selected entity, LBCC, is focused and larger than the other entities.
A screen shot of the labeling dashboard showing two tabs: 'New Tasks' and 'Labeling History.' The 'New Tasks' tab is active, and it contains a table view of a batch of 10 labeling tasks that have been chosen for the user. Each row in the task table contains the name of an entity, a drop-down select form to assign the entity's label, and the source sentence that mentions the entity. In this image, all the label selection forms have been pre-filled with Ardis AI's label recommendations for the entities. The entity and label recommendation pairs are: 'Arlington National Cemetery' - 'facility', 'Graeme McDowell' - 'person', 'Human Rights Watch' - 'organization', 'Smithsonian' - 'facility', 'Zoe' - 'person', 'Korean Empire' - 'location', 'North Waziristan' - 'location', 'Sprint' - 'organization', 'Veracruz' - 'location', 'Georg Wilhelm Friedrich Hegel' - 'person'. Below the labeling tasks label, the list of user-selected labels, the entity detail panel, and some of the entity names are also visible. The entity 'Louis Pasteur' is selected, and it has been given a recommended label of 'person.'

Take advantage of active learning to label only the highest-value tasks

We assign you batches of 10 entities with source sentences to label. Label your first batch, and then later tasks will come with recommended labels.

When you submit tasks, we update our label recommendations and choose your next batch of highest-value tasks.

Another screenshot of the labeling dashboard depicting the list of user-selected labels, the entity detail panel, and the grid of randomly selected entity names from the data set. In the list of user-selected labels, the 'event' label has focus. As a result, the opacity has been reduced for all entities in the grid below with label recommendations other than 'event.' The following 'event' entities are visible: ' Commonwealth Games', 'Six-Day War', 'Russian Revolution', 'Armistice', 'Tiananmen Square', and'First World War.'

Watch our AI adapt to your choices in real time

Submit a batch of tasks, and see the effect immediately as we update our label recommendations for hundreds of the entities in the dataset.

Add or remove labels to your label set at any time.

Start Labeling