Add summarization script

README.md

@@ -1,2 +1,19 @@
 # Codex-Test
+
 Testing Codex Features
+
+## Summarizer Script
+
+`summarizer.py` provides abstractive summarization for long-form news
+articles using a transformer-based model. Named entities are extracted
+with spaCy and the script appends any missing entities to the summary to
+help preserve important relationships.
+
+### Usage
+```
+python summarizer.py path/to/article.txt
+```
+
+Ensure that the `transformers` and `spacy` packages are installed and
+that the spaCy model `en_core_web_sm` is available. Model weights must be
+downloaded beforehand in environments without network access.

summarizer.py

@@ -0,0 +1,82 @@
+#!/usr/bin/env python3
+"""Abstractive summarization script preserving named entities.
+
+This script reads a text file containing a news article and outputs an
+abstractive summary generated by a transformer-based model from the
+Hugging Face Transformers library. It uses spaCy to extract named
+entities from the original article and checks that the generated summary
+includes these entities. Entities missing from the summary are appended
+at the end to encourage preservation of key information and
+relationships.
+
+Example usage:
+    python summarizer.py article.txt
+
+Requirements:
+    transformers
+    spacy
+    en_core_web_sm (spaCy model)
+
+Note: Running this script requires downloading model weights. In a
+restricted environment without network access, download the required
+models beforehand.
+"""
+
+import argparse
+from pathlib import Path
+from typing import Set
+
+import spacy
+from transformers import pipeline
+
+
+def load_article(path: Path) -> str:
+    """Read the article from a file."""
+    with path.open(encoding="utf-8") as f:
+        return f.read()
+
+
+def extract_entities(text: str, nlp) -> Set[str]:
+    """Return the set of named entities in the text."""
+    doc = nlp(text)
+    return {ent.text for ent in doc.ents}
+
+
+def generate_summary(text: str, model_name: str) -> str:
+    """Generate a summary using a transformer-based summarization pipeline."""
+    summarizer = pipeline("summarization", model=model_name)
+    result = summarizer(text, max_length=200, min_length=60, truncation=True)
+    return result[0]["summary_text"]
+
+
+def ensure_entities(summary: str, entities: Set[str]) -> str:
+    """Append missing named entities to the summary."""
+    missing = [ent for ent in entities if ent not in summary]
+    if missing:
+        summary += "\n\nMissing entities: " + ", ".join(missing)
+    return summary
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description="Summarize news articles")
+    parser.add_argument("article", type=Path, help="Path to the article text file")
+    parser.add_argument(
+        "--model",
+        default="facebook/bart-large-cnn",
+        help="Hugging Face model for summarization",
+    )
+    args = parser.parse_args()
+
+    text = load_article(args.article)
+
+    nlp = spacy.load("en_core_web_sm")
+    entities = extract_entities(text, nlp)
+
+    summary = generate_summary(text, args.model)
+    summary = ensure_entities(summary, entities)
+
+    print(summary)
+
+
+if __name__ == "__main__":
+    main()