feat: add date cues

changelog.md

@@ -8,6 +8,7 @@
   during the processing. Autocasting should result in a slight speedup, but may lead to numerical instability.
 - Use `torch.inference_mode` to disable view tracking and version counter bumps during inference.
 - Added a new NER pipeline for suicide attempt detection
+- Added date cues (regular expression matches that contributed to a date being detected) under the extension `ent._.date_cues`
 
 ### Changed
 

edsnlp/pipes/misc/dates/dates.py

@@ -1,4 +1,5 @@
 """`eds.dates` pipeline."""
+
 import warnings
 from itertools import chain
 from typing import Dict, Iterable, List, Optional, Tuple, Union
@@ -174,6 +175,8 @@ class DatesMatcher(BaseNERComponent):
         Label to use for periods
     span_setter : SpanSetterArg
         How to set matches in the doc.
+    explain : bool
+        Whether to keep track of regex cues for each entity.
 
     Authors and citation
     --------------------
@@ -206,10 +209,12 @@ def __init__(
             "durations": ["duration"],
             "periods": ["period"],
         },
+        explain: bool = False,
     ):
         self.date_label = date_label
         self.duration_label = duration_label
         self.period_label = period_label
+        self.explain = explain
 
         # Backward compatibility
         if as_ents is True:
@@ -302,6 +307,9 @@ def set_extensions(self) -> None:
         if not Span.has_extension(self.period_label):
             Span.set_extension(self.period_label, default=None)
 
+        if not Span.has_extension("date_cues"):
+            Span.set_extension("date_cues", default=None)
+
     def process(self, doc: Doc) -> List[Tuple[Span, Dict[str, str]]]:
         """
         Find dates in doc.
@@ -406,6 +414,9 @@ def parse(
                 span.label_ = self.duration_label
                 span._.duration = parsed
 
+            if self.explain:
+                span._.date_cues = groupdict
+
         return [span for span, _ in matches]
 
     def process_periods(self, dates: List[Span]) -> List[Span]:

tests/pipelines/misc/test_dates.py

@@ -72,7 +72,9 @@
 
 @fixture(autouse=True)
 def add_date_pipeline(blank_nlp: PipelineProtocol):
-    blank_nlp.add_pipe("eds.dates", config=dict(detect_periods=True, as_ents=True))
+    blank_nlp.add_pipe(
+        "eds.dates", config=dict(detect_periods=True, as_ents=True, explain=True)
+    )
 
 
 def test_dates_component(blank_nlp: PipelineProtocol):
@@ -89,6 +91,7 @@ def test_dates_component(blank_nlp: PipelineProtocol):
 
         for span, entity in zip(spans, entities):
             assert span.text == text[entity.start_char : entity.end_char]
+            assert bool(span._.date_cues)
 
             date = span._.date if span.label_ == "date" else span._.duration
             d = {modifier.key: modifier.value for modifier in entity.modifiers}