Update card: read feature names from model.config.id2label

README.md

@@ -1,20 +1,27 @@
 # Neurobiber: Fast and Interpretable Stylistic Feature Extraction
 
-**Neurobiber** is a transformer-based model that quickly predicts **96 interpretable stylistic features** in text. These features are inspired by Biber’s multidimensional framework of linguistic style, capturing everything from **pronouns** and **passives** to **modal verbs** and **discourse devices**. By combining a robust linguistically informed feature set with the speed of neural inference, NeuroBiber enables large-scale stylistic analyses that were previously infeasible.
+Neurobiber is a transformer-based model that quickly predicts 96 interpretable
+stylistic features in text. These features are inspired by Biber's
+multidimensional framework of linguistic style, capturing everything from
+pronouns and passives to modal verbs and discourse devices. By combining a robust
+linguistically informed feature set with the speed of neural inference, Neurobiber
+enables large-scale stylistic analyses that were previously infeasible.
 
 ## Why Neurobiber?
 
-Extracting Biber-style features typically involves running a full parser or specialized tagger, which can be computationally expensive for large datasets or real-time applications. NeuroBiber overcomes these challenges by:
-- **Operating up to 56x faster** than parsing-based approaches.
-- Retaining the **interpretability** of classical Biber-like feature definitions.
-- Delivering **high accuracy** on diverse text genres (e.g., social media, news, literary works).
-- Allowing seamless integration with **modern deep learning** pipelines via Hugging Face.
+Extracting Biber-style features typically involves running a full parser or
+specialized tagger, which can be computationally expensive for large datasets or
+real-time applications. Neurobiber overcomes these challenges by:
 
-By bridging detailed linguistic insights and industrial-scale performance, Neurobiber supports tasks in register analysis, style transfer, and more.
+- Operating up to 56x faster than parsing-based approaches.
+- Retaining the interpretability of classical Biber-like feature definitions.
+- Delivering high accuracy on diverse text genres (e.g., social media, news, literary works).
+- Allowing seamless integration with modern deep learning pipelines via Hugging Face.
 
 ## Example Script
 
-Below is an **example** showing how to load Neurobiber from Hugging Face, process single or multiple texts, and obtain a 96-dimensional binary vector for each input.
+The model now ships the feature names in its config, so you can map each output
+dimension to its feature via `model.config.id2label` - no manual feature list.
 
 ```python
 import torch
@@ -24,25 +31,6 @@ from transformers import AutoTokenizer, AutoModelForSequenceClassification
 MODEL_NAME = "Blablablab/neurobiber"
 CHUNK_SIZE = 512  # Neurobiber was trained with max_length=512
 
-# List of the 96 features that Neurobiber can predict
-BIBER_FEATURES = [
-    "BIN_QUAN","BIN_QUPR","BIN_AMP","BIN_PASS","BIN_XX0","BIN_JJ",
-    "BIN_BEMA","BIN_CAUS","BIN_CONC","BIN_COND","BIN_CONJ","BIN_CONT",
-    "BIN_DPAR","BIN_DWNT","BIN_EX","BIN_FPP1","BIN_GER","BIN_RB",
-    "BIN_PIN","BIN_INPR","BIN_TO","BIN_NEMD","BIN_OSUB","BIN_PASTP",
-    "BIN_VBD","BIN_PHC","BIN_PIRE","BIN_PLACE","BIN_POMD","BIN_PRMD",
-    "BIN_WZPRES","BIN_VPRT","BIN_PRIV","BIN_PIT","BIN_PUBV","BIN_SPP2",
-    "BIN_SMP","BIN_SERE","BIN_STPR","BIN_SUAV","BIN_SYNE","BIN_TPP3",
-    "BIN_TIME","BIN_NOMZ","BIN_BYPA","BIN_PRED","BIN_TOBJ","BIN_TSUB",
-    "BIN_THVC","BIN_NN","BIN_DEMP","BIN_DEMO","BIN_WHQU","BIN_EMPH",
-    "BIN_HDG","BIN_WZPAST","BIN_THAC","BIN_PEAS","BIN_ANDC","BIN_PRESP",
-    "BIN_PROD","BIN_SPAU","BIN_SPIN","BIN_THATD","BIN_WHOBJ","BIN_WHSUB",
-    "BIN_WHCL","BIN_ART","BIN_AUXB","BIN_CAP","BIN_SCONJ","BIN_CCONJ",
-    "BIN_DET","BIN_EMOJ","BIN_EMOT","BIN_EXCL","BIN_HASH","BIN_INF",
-    "BIN_UH","BIN_NUM","BIN_LAUGH","BIN_PRP","BIN_PREP","BIN_NNP",
-    "BIN_QUES","BIN_QUOT","BIN_AT","BIN_SBJP","BIN_URL","BIN_WH",
-    "BIN_INDA","BIN_ACCU","BIN_PGAS","BIN_CMADJ","BIN_SPADJ","BIN_X"
-]
 
 def load_model_and_tokenizer():
     tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, use_fast=True)
@@ -50,12 +38,14 @@ def load_model_and_tokenizer():
     model.eval()
     return model, tokenizer
 
+
 def chunk_text(text, chunk_size=CHUNK_SIZE):
     tokens = text.strip().split()
     if not tokens:
         return []
     return [" ".join(tokens[i:i + chunk_size]) for i in range(0, len(tokens), chunk_size)]
 
+
 def get_predictions_chunked_batch(model, tokenizer, texts, chunk_size=CHUNK_SIZE, subbatch_size=32):
     chunked_texts = []
     chunk_indices = []
@@ -104,27 +94,32 @@ def get_predictions_chunked_batch(model, tokenizer, texts, chunk_size=CHUNK_SIZE
 
     return np.array(predictions)
 
+
 def predict_batch(model, tokenizer, texts, chunk_size=CHUNK_SIZE, subbatch_size=32):
     return get_predictions_chunked_batch(model, tokenizer, texts, chunk_size, subbatch_size)
 
+
 def predict_text(model, tokenizer, text, chunk_size=CHUNK_SIZE, subbatch_size=32):
     batch_preds = predict_batch(model, tokenizer, [text], chunk_size, subbatch_size)
     return batch_preds[0]
 ```
 
 ## Single-Text Usage
-``` python
+
+```python
 model, tokenizer = load_model_and_tokenizer()
 sample_text = "This is a sample text demonstrating certain stylistic features."
 predictions = predict_text(model, tokenizer, sample_text)
-print("Binary feature vector:", predictions)
-# For example: [0, 1, 0, 1, ... 1, 0] (96-length)
 
+# Map the 96-dim binary vector to feature names straight from the model config.
+present = {model.config.id2label[i]: int(v) for i, v in enumerate(predictions)}
+print(present)                                   # {'BIN_QUAN': 0, 'BIN_QUPR': 1, ...}
+print([f for f, v in present.items() if v])      # just the detected features
 ```
 
 ## Batch Usage
-``` python
 
+```python
 docs = [
     "First text goes here.",
     "Second text, slightly different style."
@@ -132,20 +127,35 @@ docs = [
 model, tokenizer = load_model_and_tokenizer()
 preds = predict_batch(model, tokenizer, docs)
 print(preds.shape)  # (2, 96)
-```
 
+# Names for any row come from the config:
+id2label = model.config.id2label
+for row in preds:
+    print([id2label[i] for i, v in enumerate(row) if v])
+```
 
 ## How It Works
 
-Neurobiber is fine-tuned RoBERTa. Given a text:
-1. The text is split into **chunks** (up to 512 tokens each).
-2. Each chunk is fed through the model to produce **96 logistic outputs** (one per feature).
-3. The feature probabilities are aggregated across chunks so that each feature is marked as `1` if it appears in at least one chunk.
+Neurobiber is a fine-tuned RoBERTa. Given a text:
+
+1. The text is split into chunks (up to 512 tokens each).
+2. Each chunk is fed through the model to produce 96 logistic outputs (one per feature).
+3. The feature probabilities are aggregated across chunks so that each feature is
+   marked as `1` if it appears in at least one chunk.
+
+Each row in `preds` is a 96-element array. The mapping from index to feature name
+is published in `model.config.id2label` (and the reverse in `model.config.label2id`).
 
-Each row in preds is a 96-element array corresponding to the feature order in BIBER_FEATURES.
+### Interpreting Outputs
 
-Interpreting Outputs
+- Each element is a binary label (0 or 1) indicating the model's detection of a
+  specific linguistic feature (e.g., `BIN_VBD` for past tense verbs).
+- For long texts, segments of length 512 tokens are scored independently; if a
+  feature appears in any chunk, the output is `1` for that feature.
 
-- Each element in the vector is a binary label (0 or 1), indicating the model’s detection of a specific linguistic feature (e.g., BIN_VBD for past tense verbs).
-- For long texts, we chunk them into segments of length 512 tokens. If a feature appears in any chunk, you get a 1 for that feature.
+## Note on Feature Names
 
+The 96 features and their order are defined by biberplus
+(`biberplus.tagger.constants.BIBER_PLUS_TAGS`, prefixed with `BIN_`) and match the
+training label order. This mapping is embedded in the model config, so prefer
+`model.config.id2label` over any hardcoded list.