Rename WordDetection -> WordClassification (better fit for the default behavior)

By default the task does multi-class word classification; the
"detection" framing is only accurate in keyword-detection mode
(``keyword_detection="cat"`` -> binary y/n). WordClassification is the
clearer name for both pnpl.tasks.libribrain.WordClassification and
the per-dataset variants in gwilliams2022 and pallier2025.

A backwards-compatible ``WordDetection = WordClassification`` alias is
exported from each module so code written against pnpl 0.1.0
(``from pnpl.tasks import WordDetection``) keeps working unchanged.
The public-API test now asserts both names point at the same class.

Files renamed (via git mv) so import paths track the class name:
  pnpl/tasks/libribrain/word_detection.py
    -> pnpl/tasks/libribrain/word_classification.py
  pnpl/tasks/gwilliams2022/word_detection.py
    -> pnpl/tasks/gwilliams2022/word_classification.py
  pnpl/tasks/pallier2025/word_detection.py
    -> pnpl/tasks/pallier2025/word_classification.py

Internal usage in dataset compat wrappers, examples, tests, and docs
all switched to the canonical name. Stale autosummary stubs deleted;
new ones regenerate at the WordClassification path.

Author: gereonelvers

docs/api/generated/pnpl.tasks.gwilliams2022.WordClassification.rst

@@ -0,0 +1,33 @@
+pnpl.tasks.gwilliams2022.WordClassification
+===========================================
+
+.. currentmodule:: pnpl.tasks.gwilliams2022
+
+.. autoclass:: WordClassification
+
+   
+   .. automethod:: __init__
+
+   
+   .. rubric:: Methods
+
+   .. autosummary::
+   
+      ~WordClassification.__init__
+      ~WordClassification.collect_samples
+      ~WordClassification.get_label
+   
+   
+
+   
+   
+   .. rubric:: Attributes
+
+   .. autosummary::
+   
+      ~WordClassification.label_info
+      ~WordClassification.require_pronounced
+      ~WordClassification.tmax
+      ~WordClassification.tmin
+   
+   

docs/api/generated/pnpl.tasks.gwilliams2022.WordDetection.rst

@@ -0,0 +1,38 @@
+pnpl.tasks.libribrain.WordClassification
+========================================
+
+.. currentmodule:: pnpl.tasks.libribrain
+
+.. autoclass:: WordClassification
+
+   
+   .. automethod:: __init__
+
+   
+   .. rubric:: Methods
+
+   .. autosummary::
+   
+      ~WordClassification.__init__
+      ~WordClassification.collect_samples
+      ~WordClassification.get_label
+   
+   
+
+   
+   
+   .. rubric:: Attributes
+
+   .. autosummary::
+   
+      ~WordClassification.is_keyword_mode
+      ~WordClassification.keyword_detection
+      ~WordClassification.label_info
+      ~WordClassification.max_word_length
+      ~WordClassification.min_word_length
+      ~WordClassification.negative_buffer
+      ~WordClassification.positive_buffer
+      ~WordClassification.tmax
+      ~WordClassification.tmin
+   
+   

docs/api/generated/pnpl.tasks.libribrain.WordClassification.rst

@@ -0,0 +1,35 @@
+pnpl.tasks.pallier2025.WordClassification
+=========================================
+
+.. currentmodule:: pnpl.tasks.pallier2025
+
+.. autoclass:: WordClassification
+
+   
+   .. automethod:: __init__
+
+   
+   .. rubric:: Methods
+
+   .. autosummary::
+   
+      ~WordClassification.__init__
+      ~WordClassification.collect_samples
+      ~WordClassification.get_label
+   
+   
+
+   
+   
+   .. rubric:: Attributes
+
+   .. autosummary::
+   
+      ~WordClassification.keep_top_k
+      ~WordClassification.label_info
+      ~WordClassification.max_word_length
+      ~WordClassification.min_word_length
+      ~WordClassification.tmax
+      ~WordClassification.tmin
+   
+   

docs/api/generated/pnpl.tasks.libribrain.WordDetection.rst

@@ -44,12 +44,12 @@ module.
    pnpl.tasks.base.TaskProtocol
    pnpl.tasks.libribrain.SpeechDetection
    pnpl.tasks.libribrain.PhonemeClassification
-   pnpl.tasks.libribrain.WordDetection
+   pnpl.tasks.libribrain.WordClassification
    pnpl.tasks.gwilliams2022.PhonemeClassification
-   pnpl.tasks.gwilliams2022.WordDetection
+   pnpl.tasks.gwilliams2022.WordClassification
    pnpl.tasks.armeni2022.PhonemeClassification
    pnpl.tasks.schoffelen2019.TrialEpoching
-   pnpl.tasks.pallier2025.WordDetection
+   pnpl.tasks.pallier2025.WordClassification
 ```
 
 ## Preprocessing

docs/api/generated/pnpl.tasks.pallier2025.WordClassification.rst

@@ -60,7 +60,7 @@ prints a warning.
 ## Selected arguments
 
 - `task` — any object implementing `TaskProtocol`. Built-ins live in
-  `pnpl.tasks.gwilliams2022` (`PhonemeClassification`, `WordDetection`).
+  `pnpl.tasks.gwilliams2022` (`PhonemeClassification`, `WordClassification`).
 - `preprocessing` — pipeline string used in derivative filenames
   (e.g. `"notch+bp+ds"`). Set to `None` to materialize H5 from the raw
   KIT recording without filtering.
@@ -77,10 +77,10 @@ prints a warning.
 ## Available tasks
 
 ```python
-from pnpl.tasks.gwilliams2022 import PhonemeClassification, WordDetection
+from pnpl.tasks.gwilliams2022 import PhonemeClassification, WordClassification
 ```
 
 - `PhonemeClassification(tmin, tmax, label_type="phoneme" | "voicing")`
   — sample windowed around each phoneme onset.
-- `WordDetection(tmin, tmax, require_pronounced=True)` — sample
+- `WordClassification(tmin, tmax, require_pronounced=True)` — sample
   windowed around each word onset; label is the word string.

docs/api/generated/pnpl.tasks.pallier2025.WordDetection.rst

@@ -23,11 +23,11 @@ The companion paper is d'Ascoli, Bel, Rapin, King et al.,
 
 ```python
 from pnpl.datasets import Pallier2025
-from pnpl.tasks.pallier2025 import WordDetection
+from pnpl.tasks.pallier2025 import WordClassification
 
 ds = Pallier2025(
     data_path="./data/pallier2025",
-    task=WordDetection(tmin=0.0, tmax=3.0),
+    task=WordClassification(tmin=0.0, tmax=3.0),
     include_subjects=["01"],
     include_runs=["01"],            # one ~10 min audiobook segment
     preprocessing="notch+bp+ds",    # 50/100 Hz notch, 0.1–125 Hz bp, 250 Hz resample
@@ -76,7 +76,7 @@ notch or SSS. The pipeline is configurable on a per-step basis:
 ```python
 ds = Pallier2025(
     data_path="./data/pallier2025",
-    task=WordDetection(tmin=0.0, tmax=3.0),
+    task=WordClassification(tmin=0.0, tmax=3.0),
     include_subjects=["01"],
     preprocessing="bp+ds",
     preprocessing_config={
@@ -91,7 +91,7 @@ mechanism (defaults < JSON < dataset config).
 
 ## Selected arguments
 
-- `task` — currently `pnpl.tasks.pallier2025.WordDetection`.
+- `task` — currently `pnpl.tasks.pallier2025.WordClassification`.
 - `preprocessing` — pipeline string used in derivative filenames.
   Default `"notch+bp+ds"`. `None` materializes H5 from the raw FIF
   unchanged.
@@ -106,7 +106,7 @@ mechanism (defaults < JSON < dataset config).
 
 ## Available task
 
-`WordDetection(tmin, tmax, min_word_length, max_word_length, keep_top_k)` —
+`WordClassification(tmin, tmax, min_word_length, max_word_length, keep_top_k)` —
 windowed around each word onset.
 
 - `tmin`, `tmax` — defaults `0.0`, `3.0` to match d'Ascoli 2025.

docs/api/index.md

@@ -53,7 +53,7 @@ default `label_info` implementation if you set `_classes` /
 from pnpl.tasks import (
     SpeechDetection,
     PhonemeClassification,
-    WordDetection,
+    WordClassification,
 )
 ```
 
@@ -65,20 +65,20 @@ from pnpl.tasks import (
   *speech* / *silence*. Returns a per-time-point label array.
 - `PhonemeClassification(tmin, tmax, label_type="phoneme" | "voicing", exclude_phonemes=[])`
   — sample windowed around each phoneme onset.
-- `WordDetection(tmin, tmax, min_word_length=1, max_word_length=None, keyword_detection=None)`
+- `WordClassification(tmin, tmax, min_word_length=1, max_word_length=None, keyword_detection=None)`
   — multi-class word classification *or* binary keyword detection
   (`keyword_detection="cat"` → 1 if the window's word is `"cat"`, else
   0). `tmin` / `tmax` may be `None` to auto-compute from word duration.
 
 ## MEG-MASC tasks (Gwilliams 2022)
 
 ```python
-from pnpl.tasks.gwilliams2022 import PhonemeClassification, WordDetection
+from pnpl.tasks.gwilliams2022 import PhonemeClassification, WordClassification
 ```
 
 - `PhonemeClassification(tmin, tmax, label_type="phoneme" | "voicing")`
   — phoneme-aligned epochs from the MEG-MASC events.tsv.
-- `WordDetection(tmin, tmax, require_pronounced=True)` — word-aligned
+- `WordClassification(tmin, tmax, require_pronounced=True)` — word-aligned
   epochs; label is the lower-cased word string.
 
 ## Armeni 2022 tasks
@@ -94,10 +94,10 @@ from pnpl.tasks.armeni2022 import PhonemeClassification
 ## Pallier 2025 (LittlePrince Listen) tasks
 
 ```python
-from pnpl.tasks.pallier2025 import WordDetection
+from pnpl.tasks.pallier2025 import WordClassification
 ```
 
-- `WordDetection(tmin, tmax, min_word_length, max_word_length, keep_top_k)`
+- `WordClassification(tmin, tmax, min_word_length, max_word_length, keep_top_k)`
   — windowed around each word onset (`trial_type='Word'` rows in the
   events.tsv). Default `tmin=0.0`, `tmax=3.0` matches the d'Ascoli et
   al. (Nat Commun 2025) recipe. `keep_top_k` restricts the label