## Files

*   **training-set-100000.csv** - the training set containing exactly 100,000 annotated articles. 
*   **test-set-20000.csv** - the test set containing exactly 20,000 annotated articles for you to try to evaluate your model for tuning and other kinds of improvements.
*  **judge-set-10001-unannotated.csv** - the judge set containing exactly 10,001 unannoated articles to have your best model predict the MeSH terms for each and prepare for submission.
*   **sample_submission.csv** - a sample submission file in the correct format. Please note the `meshMajor` column. In the sample, the list is purely random.

## The training dataset for your consumption
The data set for your reference contains annotated articles from [PubMed at https://pubmed.ncbi.nlm.nih.gov](https://pubmed.ncbi.nlm.nih.gov), where `annotated` means that the `Medical Subject Headings` (`MeSH`) terms have been assigned to the articles by the human curators at PubMed.

The dataset is served as a CSV file with 1000,000 articles as rows, each having information about the articles as 4 columns:  
1. `pmid ` - A PMID, or PubMed Identifier, is a unique number assigned to every citation in the PubMed database, which is maintained by the National Library of Medicine. Data type: integer.
2. `title ` - The title of the article. Data type: string.
3. `abstractText` - the abstract of the article. Data type is string. Example: "AIMS: In light of recently found contribution ..."
4. `meshMajor`- list of comma separated annotated MeSH labels for the article. Data type is string. Example:  "['Adult', 'Aged', ...]"

The training set contains exactly `100,000` articles covering `22,373` MeSH terms.

## Annotated Test set
The test set contains exactly `20,000` articles covering `16,247` MeSH. Structure of the CSV file is exactly the same as the training set. Please note, the articles in the test set do not introduce any new MeSH terms.

## Unannotated Judge set
The judge set will consist of articles having `almost` similar format as the test set, except the `meshMajor` keys are absent for each article. The judge set contains exactly `10,001` unannoated PubMed articles.