a b/docs/utilities/connectors/brat.md 

  1
  
# BRAT Connector





  
  

  2
  





  
  

  3
  
BRAT is currently the only supported in-text annotation editor at EDS. BRAT annotations are in the [standoff format](https://brat.nlplab.org/standoff.html). Consider the following document:





  
  

  4
  





  
  

  5
  
```





  
  

  6
  
Le patient est admis pour une pneumopathie au coronavirus.





  
  

  7
  
On lui prescrit du paracétamol.





  
  

  8
  
```





  
  

  9
  





  
  

  10
  
It could be annotated as follows :





  
  

  11
  





  
  

  12
  
```





  
  

  13
  
T1  Patient 4 11    patient





  
  

  14
  
T2  Disease 31 58   pneumopathie au coronavirus





  
  

  15
  
T3  Drug 79 90  paracétamol





  
  

  16
  
```





  
  

  17
  





  
  

  18
  
The point of the BRAT connector is to go from the standoff annotation format to an annotated spaCy document :





  
  

  19
  





  
  

  20
  
```{ .python .no-check }





  
  

  21
  
import edsnlp





  
  

  22
  
from edsnlp.connectors.brat import BratConnector





  
  

  23
  





  
  

  24
  
# Instantiate the connector





  
  

  25
  
brat = BratConnector("path/to/brat")





  
  

  26
  





  
  

  27
  
# Instantiate the spacy pipeline





  
  

  28
  
nlp = edsnlp.blank("eds")





  
  

  29
  





  
  

  30
  
# Convert all BRAT files to a list of documents





  
  

  31
  
docs = brat.brat2docs(nlp)





  
  

  32
  
doc = docs[0]





  
  

  33
  





  
  

  34
  
doc.ents





  
  

  35
  
# Out: [patient, pneumopathie au coronavirus, paracétamol]





  
  

  36
  





  
  

  37
  
doc.ents[0].label_





  
  

  38
  
# Out: Patient





  
  

  39
  
```





  
  

  40
  





  
  

  41
  
The connector can also go the other way around, enabling pre-annotations and an ersatz of active learning.