|
a |
|
b/edsnlp/connectors/brat.py |
|
|
1 |
import warnings |
|
|
2 |
from pathlib import Path |
|
|
3 |
from typing import List, Optional, Union |
|
|
4 |
|
|
|
5 |
from spacy.tokens import Doc |
|
|
6 |
|
|
|
7 |
from edsnlp.core import PipelineProtocol |
|
|
8 |
from edsnlp.data.converters import AttributesMappingArg |
|
|
9 |
from edsnlp.data.standoff import ( |
|
|
10 |
dump_standoff_file, |
|
|
11 |
parse_standoff_file, |
|
|
12 |
read_standoff, |
|
|
13 |
write_standoff, |
|
|
14 |
) |
|
|
15 |
from edsnlp.utils.span_getters import ( |
|
|
16 |
SpanSetterArg, |
|
|
17 |
validate_span_setter, |
|
|
18 |
) |
|
|
19 |
|
|
|
20 |
|
|
|
21 |
class BratConnector(object): |
|
|
22 |
""" |
|
|
23 |
Deprecated. Use `edsnlp.data.read_standoff` and `edsnlp.data.write_standoff` |
|
|
24 |
instead. |
|
|
25 |
Two-way connector with BRAT. Supports entities only. |
|
|
26 |
|
|
|
27 |
Parameters |
|
|
28 |
---------- |
|
|
29 |
directory : Union[str, Path] |
|
|
30 |
Directory containing the BRAT files. |
|
|
31 |
n_jobs : int, optional |
|
|
32 |
Number of jobs for multiprocessing, by default 1 |
|
|
33 |
attributes: Optional[Union[Sequence[str], Mapping[str, str]]] |
|
|
34 |
Mapping from BRAT attributes to spaCy Span extensions. |
|
|
35 |
Extensions / attributes that are not in the mapping are not imported or exported |
|
|
36 |
If left to None, the mapping is filled with all BRAT attributes. |
|
|
37 |
span_groups: Optional[Sequence[str]] |
|
|
38 |
Additional span groups to look for entities in spaCy documents when exporting. |
|
|
39 |
Missing label (resp. span group) names are not imported (resp. exported) |
|
|
40 |
If left to None, the sequence is filled with all BRAT entity labels. |
|
|
41 |
""" |
|
|
42 |
|
|
|
43 |
def __init__( |
|
|
44 |
self, |
|
|
45 |
directory: Union[str, Path], |
|
|
46 |
n_jobs: int = 1, |
|
|
47 |
attributes: Optional[AttributesMappingArg] = None, |
|
|
48 |
bool_attributes: Optional[List[str]] = [], |
|
|
49 |
span_groups: SpanSetterArg = ["ents", "*"], |
|
|
50 |
keep_raw_attribute_values: bool = False, |
|
|
51 |
): |
|
|
52 |
warnings.warn( |
|
|
53 |
"This connector is deprecated and will be removed in a future version.\n" |
|
|
54 |
"Use `edsnlp.data.read_standoff` and `edsnlp.data.write_standoff` instead.", |
|
|
55 |
DeprecationWarning, |
|
|
56 |
) |
|
|
57 |
self.directory: Path = Path(directory) |
|
|
58 |
self.attr_map = attributes |
|
|
59 |
self.span_setter = validate_span_setter(span_groups) |
|
|
60 |
self.keep_raw_attribute_values = keep_raw_attribute_values |
|
|
61 |
self.bool_attributes = list(bool_attributes) |
|
|
62 |
|
|
|
63 |
def brat2docs(self, nlp: PipelineProtocol, run_pipe=False) -> List[Doc]: |
|
|
64 |
res = read_standoff( |
|
|
65 |
path=self.directory, |
|
|
66 |
nlp=nlp, |
|
|
67 |
keep_txt_only_docs=True, |
|
|
68 |
span_attributes=self.attr_map, |
|
|
69 |
span_setter=self.span_setter, |
|
|
70 |
keep_raw_attribute_values=self.keep_raw_attribute_values, |
|
|
71 |
bool_attributes=self.bool_attributes, |
|
|
72 |
) |
|
|
73 |
return list(nlp.pipe(res) if run_pipe else res) |
|
|
74 |
|
|
|
75 |
def docs2brat(self, docs: List[Doc]) -> None: |
|
|
76 |
""" |
|
|
77 |
Writes a list of spaCy documents to file. |
|
|
78 |
""" |
|
|
79 |
write_standoff( |
|
|
80 |
docs, |
|
|
81 |
self.directory, |
|
|
82 |
span_getter=self.span_setter, |
|
|
83 |
span_attributes=self.attr_map or {}, |
|
|
84 |
overwrite=True, |
|
|
85 |
) |
|
|
86 |
|
|
|
87 |
|
|
|
88 |
load_from_brat = parse_standoff_file |
|
|
89 |
export_to_brat = dump_standoff_file |