# Coleridge Initiative - Show US the Data — Data

来源：https://www.kaggle.com/competitions/coleridgeinitiative-show-us-the-data/data

采集时间：2026-10-25T10:32:27Z

The objective of the competition is to identify the mention of datasets within scientific publications. Your predictions will be short excerpts from the publications that appear to note a dataset. Predictions that more accurately match the precise words used to identify the dataset within the publication will score higher. Predictions should be cleaned using the `clean_text` function from the [Evaluation page](https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/overview/evaluation) to ensure proper matching.

Publications are provided in JSON format, broken up into sections with section titles.

The goal in this competition is not just to match known dataset strings but to generalize to datasets that have never been seen before using NLP and statistical techniques. A percentage of the public test set publications are drawn from the training set - not all datasets have been identified in train, so these unidentified datasets have been used as a portion of the public test labels. These should serve as guides for the difficult task of labeling the private test set.

Note that the hidden test set has roughly ~8000 publications, many times the size of the public test set. Plan your compute time accordingly.

## Files

*   **train** - the full text of the training set's publications in JSON format, broken into sections with section titles
*   **test** - the full text of the test set's publications in JSON format, broken into sections with section titles
*   **train.csv** - labels and metadata for the training set
*   **sample_submission.csv** - a sample submission file in the correct format

## Columns

*   `id` - publication `id` - note that there are multiple rows for some training documents, indicating multiple mentioned datasets
*   `pub_title` - title of the publication (a small number of publications have the same title)
*   `dataset_title` - the title of the dataset that is mentioned within the publication
*   `dataset_label` - a portion of the text that indicates the dataset
*   `cleaned_label` - the `dataset_label`, as passed through the `clean_text` function from the [Evaluation page](https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/overview/evaluation)