Update evaluation suite

pyproject.toml

@@ -11,7 +11,6 @@ dependencies = [
     "matplotlib>=3.10.8",
     "scipy>=1.17.0",
     "seaborn>=0.13.2",
-    "sentence-transformers>=5.3.0",
 ]
 
 [build-system]
@@ -102,8 +101,6 @@ module = [
     "seaborn.*",
     "matplotlib",
     "matplotlib.*",
-    "sentence_transformers",
-    "sentence_transformers.*"
 ]
 ignore_missing_imports = true
 

src/voice/__init__.py

@@ -6,8 +6,8 @@
 """
 
 from voice.comparison import (
-    make_embedding_comparison,
-    make_stylometric_comparison,
+    make_comparison,
+    stylometric_distribution,
 )
 from voice.datasets import DatasetSpec, get_dataset
 from voice.stylometry import get_groups, get_metrics
@@ -17,6 +17,6 @@
     "get_metrics",
     "get_groups",
     "get_dataset",
-    "make_embedding_comparison",
-    "make_stylometric_comparison",
+    "stylometric_distribution",
+    "make_comparison",
 ]

src/voice/stylometry/_defaults.py → src/voice/_defaults.py

@@ -1,7 +1,7 @@
 """
-Defaults for the stylometry module.
+Defaults for the VOICE project.
 
-This module contains defaults for stylometric analysis
+This module contains defaults for the VOICE project
 e.g. default parameters for parameterised metrics
 """
 
@@ -14,19 +14,15 @@ class MetricGroup(str, Enum):
     Enumeration of stylometric metric groups.
 
     Metrics within the same group are expected to be correlated;
-    the group is the unit of analysis for Bonferroni correction.
+    groups are the unit of aggregation for the alignment score.
 
     .. attribute :: WORD_LENGTH_DISTRIBUTION
 
         Moments of the word length distribution (mean, std, skew, kurtosis).
 
-    .. attribute :: LEXICAL_RICHNESS
+    .. attribute :: VOCABULARY_RICHNESS
 
-        Type-token ratio and its moving-average variant.
-
-    .. attribute :: LEGOMENA
-
-        Hapax, dis and tri legomena ratios.
+        Type-token ratio, its moving-average variant, and legomena ratios.
 
     .. attribute :: FUNCTION_WORDS
 
@@ -42,8 +38,7 @@ class MetricGroup(str, Enum):
     """
 
     WORD_LENGTH_DISTRIBUTION = "word_length_distribution"
-    LEXICAL_RICHNESS = "lexical_richness"
-    LEGOMENA = "legomena"
+    VOCABULARY_RICHNESS = "vocabulary_richness"
     FUNCTION_WORDS = "function_words"
     CHAR_NGRAM_DIVERSITY = "char_ngram_diversity"
     TEXT_LENGTH = "text_length"
@@ -116,22 +111,6 @@ class CalibrationDefaults:
 CALIBRATION_DEFAULTS: CalibrationDefaults = CalibrationDefaults()
 
 
-@dataclass(frozen=True)
-class ComparisonDefaults:
-    """
-    Default parameters for comparing stylometric distributions.
-
-    .. attribute :: alpha
-
-        Significance level for hypothesis testing.
-    """
-
-    alpha: float = 0.05
-
-
-COMPARISON_DEFAULTS: ComparisonDefaults = ComparisonDefaults()
-
-
 @dataclass(frozen=True)
 class PlottingDefaults:
     """

src/voice/comparison/__init__.py

@@ -5,14 +5,12 @@
 - Functions for comparing distributions of stylometric metrics
 """
 
-from voice.comparison.embedding_comparison import make_embedding_comparison
-from voice.comparison.stylometric_comparison import (
-    make_stylometric_comparison,
+from voice.comparison.comparison import (
+    make_comparison,
     stylometric_distribution,
 )
 
 __all__: list[str] = [
-    "make_stylometric_comparison",
-    "make_embedding_comparison",
     "stylometric_distribution",
+    "make_comparison",
 ]

src/voice/comparison/_utils.py

@@ -10,7 +10,7 @@
 
 import numpy as np
 
-from voice.stylometry._defaults import CALIBRATION_DEFAULTS
+from voice._defaults import CALIBRATION_DEFAULTS
 
 # -----------------------------------------------------------------------------
 # Calibrated percentile