feat: add sklearn transformer

pyproject.toml

@@ -71,7 +71,8 @@ tutorials = [
     "jupyter",
     "seaborn",
     "matplotlib",
-    "datasets>=2.8.0,<2.11.0",
+    "datasets>=2.8.0,
+    "scikit-learn">=1.1.0,
 ]
 
 [project.readme]

src/textdescriptives/sklearn_featurizer.py

@@ -0,0 +1,110 @@
+from typing import Iterable, List, Optional
+
+import numpy as np
+import pandas as pd
+from sklearn.base import BaseEstimator, TransformerMixin
+from wasabi import msg
+
+from textdescriptives import extract_metrics
+
+
+def get_feature_names_from_metrics_and_model(
+    lang: Optional[str],
+    metrics: Optional[Iterable[str]],
+    spacy_model: Optional[str],
+    spacy_model_size: str,
+) -> List[str]:
+    """Get the names of the extracted features from the specified metrics
+    and model. Does this by extracting the metrics from an empty dummy text."""
+    df = extract_metrics(
+        text="",
+        lang=lang,
+        metrics=metrics,
+        spacy_model=spacy_model,
+        spacy_model_size=spacy_model_size,
+    )
+    return list(df.drop("text", axis=1).columns)
+
+
+class TextDescriptivesFeaturizer(TransformerMixin, BaseEstimator):
+    """Wrapper for extracting text metrics using textdescriptives and
+    using it in a sklearn pipeline."""
+
+    def __init__(
+        self,
+        lang: Optional[str] = None,
+        metrics: Optional[Iterable[str]] = None,
+        spacy_model: Optional[str] = None,
+        spacy_model_size: str = "lg",
+    ):
+        """Initialise the transformer with arguments to
+        textdescriptives.extract_metrics.
+
+        Args:
+            text (Union[str, List[str]]): A text or a list of texts.
+            lang (str, optional): Language of the text. If lang is set and no spacy
+                model is provided, will automatically download and use a spacy
+                model for the language. Defaults to None.
+            metrics (List[str]): Which metrics to extract.
+                One or more of ["descriptive_stats", "readability",
+                "dependency_distance", "pos_proportions", "coherence", "quality",
+                "information_theory"]. If None, will extract all metrics from
+                textdescriptives. Defaults to None.
+            spacy_model (str, optional): The spacy model to use. If not set,
+                will download one based on lang. Defaults to None.
+            spacy_model_size (str, optional): Size of the spacy model to download.
+        """
+        self.lang = lang
+        if isinstance(metrics, str):
+            metrics = [metrics]
+        self.metrics = metrics
+        self.spacy_model = spacy_model
+        self.spacy_model_size = spacy_model_size
+
+        if spacy_model is None and lang is None:
+            raise ValueError("Either a spacy model or a language must be provided.")
+        if spacy_model is not None and lang is not None:
+            msg.info(
+                "Both a spacy model and a language were provided. "
+                + "Will use the spacy model and ignore language.",
+            )
+        self.feature_names = get_feature_names_from_metrics_and_model(
+            lang=self.lang,
+            metrics=self.metrics,
+            spacy_model=self.spacy_model,
+            spacy_model_size=self.spacy_model_size,
+        )
+
+    def fit(self, X, y=None):
+        """Fit the transformer to the data. This is not needed for this
+        transformer, but is required for sklearn compatibility."""
+        return self
+
+    def transform(self, X) -> pd.DataFrame:
+        """Transform the data using textdescriptives.
+
+        Args:
+            X: Iterable of strings.
+
+        Returns:
+            Numpy array of shape (n_samples, n_features).
+        """
+        metrics = extract_metrics(
+            X,
+            lang=self.lang,
+            metrics=self.metrics,
+            spacy_model=self.spacy_model,
+            spacy_model_size=self.spacy_model_size,
+        )
+        return metrics.drop("text", axis=1)
+
+    def get_feature_names(self) -> List[str]:
+        """Get the names of the extracted features."""
+        return self.feature_names
+
+    def get_feature_names_out(self, input_features=None) -> List[str]:
+        """Get the names of the extracted features. input_features is only
+        present for API compatibility with sklearn."""
+        return self.feature_names
+
+