| import spacy | |
| class PiiSpacyAnalyzer: | |
| def __init__(self, model_name="en_core_web_lg"): | |
| """ | |
| Initializes the SpaCy model. | |
| Loads the 'large' model by default for better accuracy. | |
| """ | |
| self.nlp = None | |
| self.available = False | |
| try: | |
| print(f"⏳ Loading SpaCy model: {model_name}...") | |
| self.nlp = spacy.load(model_name) | |
| self.available = True | |
| print(f"✅ SpaCy model '{model_name}' loaded successfully.") | |
| except OSError: | |
| print(f"❌ Error: Model '{model_name}' not found.") | |
| print("👉 Run: python -m spacy download en_core_web_lg") | |
| # Optional: Fallback to small model if large fails | |
| try: | |
| self.nlp = spacy.load("en_core_web_sm") | |
| self.available = True | |
| print("⚠️ Loaded fallback model 'en_core_web_sm'.") | |
| except: | |
| self.available = False | |
| def scan(self, text: str) -> list: | |
| """ | |
| Scans text for Named Entities (Person, Org, GPE). | |
| Returns a list of dictionaries compatible with backend.py. | |
| """ | |
| if not self.available or not text: | |
| return [] | |
| # Prevent errors on huge texts by increasing limit temporarily | |
| self.nlp.max_length = len(text) + 1000 | |
| doc = self.nlp(text) | |
| detections = [] | |
| # Map SpaCy labels to your App's standard labels | |
| label_map = { | |
| "PERSON": "FIRST_NAME", | |
| "GPE": "LOCATION", # Countries, Cities, States | |
| "ORG": "ORG" # Companies, Agencies, Institutions | |
| } | |
| for ent in doc.ents: | |
| if ent.label_ in label_map: | |
| detections.append({ | |
| "label": label_map[ent.label_], | |
| "text": ent.text, | |
| "start": ent.start_char, | |
| "end": ent.end_char | |
| }) | |
| return detections | |
| def retrain(self, training_data): | |
| """ | |
| Placeholder for future dynamic training logic. | |
| You can implement the training loop here later without touching backend.py. | |
| """ | |
| pass |