diff --git a/edsnlp/pipes/trainable/embeddings/span_pooler/span_pooler.py b/edsnlp/pipes/trainable/embeddings/span_pooler/span_pooler.py
index 52d5e163b..5d2ed02e9 100644
--- a/edsnlp/pipes/trainable/embeddings/span_pooler/span_pooler.py
+++ b/edsnlp/pipes/trainable/embeddings/span_pooler/span_pooler.py
@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+import warnings
 from typing import (
     Any,
     Dict,
@@ -9,6 +10,7 @@
 
 import foldedtensor as ft
 import torch
+from confit import VisibleDeprecationWarning
 from spacy.tokens import Doc, Span
 from typing_extensions import Literal, TypedDict
 
@@ -81,11 +83,19 @@ def __init__(
         span_getter: Any = None,
     ):
         if span_getter is not None:
-            raise ValueError(
+            warnings.warn(
                 "The `span_getter` parameter of the `eds.span_pooler` component is "
                 "deprecated. Please use the `span_getter` parameter of the "
-                "`eds.span_classifier` or `eds.span_linker` components instead."
+                "`eds.span_classifier` or `eds.span_linker` components instead.",
+                VisibleDeprecationWarning,
             )
+        sub_span_getter = getattr(embedding, "span_getter", None)
+        if sub_span_getter is not None and span_getter is None:  # pragma: no cover
+            self.span_getter = sub_span_getter
+        sub_context_getter = getattr(embedding, "context_getter", None)
+        if sub_context_getter is not None:  # pragma: no cover
+            self.context_getter = sub_context_getter
+
         self.output_size = embedding.output_size if hidden_size is None else hidden_size
 
         super().__init__(nlp, name)
diff --git a/edsnlp/pipes/trainable/embeddings/text_cnn/text_cnn.py b/edsnlp/pipes/trainable/embeddings/text_cnn/text_cnn.py
index 6c46847b6..7a1a3a39a 100644
--- a/edsnlp/pipes/trainable/embeddings/text_cnn/text_cnn.py
+++ b/edsnlp/pipes/trainable/embeddings/text_cnn/text_cnn.py
@@ -63,6 +63,13 @@ def __init__(
         residual: bool = True,
         normalize: Literal["pre", "post", "none"] = "pre",
     ):
+        sub_span_getter = getattr(embedding, "span_getter", None)
+        if sub_span_getter is not None:  # pragma: no cover
+            self.span_getter = sub_span_getter
+        sub_context_getter = getattr(embedding, "context_getter", None)
+        if sub_context_getter is not None:  # pragma: no cover
+            self.context_getter = sub_context_getter
+
         super().__init__(nlp, name)
         self.embedding = embedding
         self.output_size = output_size or embedding.output_size
diff --git a/edsnlp/pipes/trainable/embeddings/transformer/transformer.py b/edsnlp/pipes/trainable/embeddings/transformer/transformer.py
index da6d19c76..4d1b68efb 100644
--- a/edsnlp/pipes/trainable/embeddings/transformer/transformer.py
+++ b/edsnlp/pipes/trainable/embeddings/transformer/transformer.py
@@ -1,3 +1,4 @@
+import warnings
 from pathlib import Path
 from typing import List, Optional, Set, Tuple, Union
 
@@ -5,7 +6,7 @@
 import tokenizers
 import tokenizers.normalizers
 import torch
-from confit import validate_arguments
+from confit import VisibleDeprecationWarning, validate_arguments
 from transformers import AutoModel, AutoTokenizer
 from transformers import BitsAndBytesConfig as BitsAndBytesConfig_
 from typing_extensions import Literal, TypedDict
@@ -148,10 +149,11 @@ def __init__(
         super().__init__(nlp, name)
 
         if span_getter is not None:
-            raise ValueError(
+            warnings.warn(
                 "The `span_getter` parameter of the `eds.transformer` component is "
                 "deprecated. Please use the `context_getter` parameter of the "
-                "other higher level task components instead."
+                "other higher level task components instead.",
+                VisibleDeprecationWarning,
             )
         self.transformer = AutoModel.from_pretrained(
             model,
diff --git a/edsnlp/pipes/trainable/ner_crf/ner_crf.py b/edsnlp/pipes/trainable/ner_crf/ner_crf.py
index 19ae1af27..ac2d19800 100644
--- a/edsnlp/pipes/trainable/ner_crf/ner_crf.py
+++ b/edsnlp/pipes/trainable/ner_crf/ner_crf.py
@@ -183,6 +183,11 @@ def __init__(
                 "You cannot set both the `labels` key of the `target_span_getter` "
                 "parameter and the `labels` parameter."
             )
+        sub_context_getter = getattr(embedding, "context_getter", None)
+        if (
+            sub_context_getter is not None and context_getter is None
+        ):  # pragma: no cover
+            context_getter = sub_context_getter
 
         super().__init__(
             nlp=nlp,
diff --git a/edsnlp/pipes/trainable/span_classifier/span_classifier.py b/edsnlp/pipes/trainable/span_classifier/span_classifier.py
index a9f66e769..b7111cbe2 100644
--- a/edsnlp/pipes/trainable/span_classifier/span_classifier.py
+++ b/edsnlp/pipes/trainable/span_classifier/span_classifier.py
@@ -212,6 +212,17 @@ def __init__(
             )
             assert attributes is None
             attributes = qualifiers
+        sub_span_getter = getattr(embedding, "span_getter", None)
+        if (
+            sub_span_getter is not None and span_getter is None
+        ):  # pragma: no cover  # noqa: E501
+            span_getter = sub_span_getter
+        sub_context_getter = getattr(embedding, "context_getter", None)
+        if (
+            sub_context_getter is not None and context_getter is None
+        ):  # pragma: no cover
+            context_getter = sub_context_getter
+
         self.values = values
         self.keep_none = keep_none
         self.bindings: List[Tuple[str, List[str], List[Any]]] = [
diff --git a/edsnlp/pipes/trainable/span_linker/span_linker.py b/edsnlp/pipes/trainable/span_linker/span_linker.py
index 46cfa2498..f73593915 100644
--- a/edsnlp/pipes/trainable/span_linker/span_linker.py
+++ b/edsnlp/pipes/trainable/span_linker/span_linker.py
@@ -286,6 +286,15 @@ def __init__(
     ):
         self.attribute = attribute
 
+        sub_span_getter = getattr(embedding, "span_getter", None)
+        if sub_span_getter is not None and span_getter is None:  # pragma: no cover
+            span_getter = sub_span_getter
+        sub_context_getter = getattr(embedding, "context_getter", None)
+        if (
+            sub_context_getter is not None and context_getter is None
+        ):  # pragma: no cover
+            context_getter = sub_context_getter
+
         super().__init__(
             nlp,
             name,
diff --git a/edsnlp/utils/span_getters.py b/edsnlp/utils/span_getters.py
index 80e6accd1..6a39c3332 100644
--- a/edsnlp/utils/span_getters.py
+++ b/edsnlp/utils/span_getters.py
@@ -273,11 +273,16 @@ def __init__(
         self,
         context_words: NonNegativeInt = 0,
         context_sents: Optional[NonNegativeInt] = 1,
+        span_getter: Optional[SpanGetterArg] = None,
     ):
         self.context_words = context_words
         self.context_sents = context_sents
+        self.span_getter = validate_span_getter(span_getter, optional=True)
+
+    def __call__(self, span: Union[Doc, Span]) -> Union[Span, List[Span]]:
+        if isinstance(span, Doc):  # pragma: no cover
+            return [self(s) for s in get_spans(span, self.span_getter)]
 
-    def __call__(self, span: Span) -> List[Span]:
         n_sents: int = self.context_sents
         n_words = self.context_words