This repository contains curation data and the target articles for Pathogen Identifier and Strain Tagger datasets.
The curation data is stored in JSON format under curation_data and divided into two subfolders: pathogens for Pathogen Identifier dataset and strains for Strain Tagger dataset.
The JSON files contain curated article hash as a key followed by IOB-format compatible tag/location key-value pairs. The tag locations are indicated inside an array on word-level as a start index, followed by a plus sign and another number that indicates the length of the tag: e.g. strains/purple: [42+2, 153+1] means that there are two strains/purple tags starting from word number 43 and 154, and spanning for two words and one word (i.e. that word only).
See TEA repository for usage example.
| Entity | Number of tags |
|---|---|
| commensals | 25 |
| negatives | 9322 |
| opportunistics | 37 |
| pathogens | 859 |
| probiotics | 20 |
| strains | 123 |
| Entity | Number of tags |
|---|---|
| negatives | 13181 |
| species | 837 |
| strains | 1788 |
The curated data is provided under MIT licence.
The target articles texts are found under source_articles. The article texts are licenced under various Creative Commons licences (i.e. BY/NC/SA).
See attribution.txt for article-level details and general attribution.