גרסת cli משולבת sklean

יצירת גרסת cli של התוכנה שמשלבת זיהוי סוג מחרוזת (זמר או אחר) לצורך שכבת זיהוי נוספת על מודל ה-NER

.github/workflows/‏‏publish-cli-ml.yml

@@ -21,7 +21,7 @@ jobs:
 
     - name: Install dependencies
       run: |
-        pip install pyinstaller music_tag jibrish_to_hebrew spacy==3.7.5
+        pip install pyinstaller music_tag jibrish_to_hebrew spacy==3.7.5 scikit-learn
 
     - name: Build EXE
       run: |

machine-learn/music_classification/model_creation/try_model.py

@@ -0,0 +1,21 @@
+import pickle
+
+# מיפוי תוויות מספריות לשמות קטגוריות
+label_mapping = {0: "ARTIST", 1: "ALBUM", 2: "SONG", 3: "RANDOM"}
+
+# טעינת המודל
+with open('music_classifier.pkl', 'rb') as f:
+    loaded_model = pickle.load(f)
+
+# פונקציה לחיזוי על מחרוזת טקסט בודדת
+def classify_text(text):
+    # המודל מצפה לקבל רשימה של טקסטים גם אם יש טקסט אחד בלבד
+    prediction = loaded_model.predict([text])
+    return prediction[0]  # החיזוי הוא רשימה ולכן אנו מחזירים את הערך הראשון
+
+# הדגמת השימוש בפונקציה
+text_input = input("הכנס מחרוזת טקסט לסיווג: ")
+prediction = classify_text(text_input)
+
+# הצגת התוצאה
+print(f"הקטגוריה של הטקסט היא: {label_mapping[prediction]} (זיהוי מספרי: {prediction})")

src/core/singles_sorter_v5.py

@@ -1,5 +1,5 @@
 # -*- coding: utf-8 -*-
-__VERSION__ = '13.6'
+__VERSION__ = '14.0'
 
 import os
 import re
@@ -15,6 +15,8 @@
 import datetime
 
 from spacy import load
+import pickle
+from sklearn import metrics
 
 class MusicSorter:
 
@@ -51,15 +53,73 @@ def __init__(self, source_dir, target_dir=None, copy_mode=False, abc_sort=False,
             self.logger.error(f"Failed to load NER model: {str(e)}")
             self.nlp = None
 
+        # Load the sklearn model
+        try:
+            with open(r'C:\Users\משתמש\Documents\GitHub\Singles-Sorter-ml\machine-learn\music_classification\model_creation\music_classifier.pkl', 'rb') as model_file:
+                self.sklearn_model = pickle.load(model_file)
+            self.logger.info("Loaded sklearn model successfully")
+        except Exception as e:
+            self.logger.error(f"Failed to load sklearn model: {str(e)}")
+            self.sklearn_model = None
+
+
+
+    def verify_artist_with_sklearn(self, artist_name):
+        if not self.sklearn_model:
+            self.logger.warning("sklearn model not available for verification")
+            return True  # Assume it's an artist if model is not available
+
+        # Prepare the input for the model
+        input_data = [artist_name]
+
+        try:
+            # Make prediction
+            prediction = self.sklearn_model.predict(input_data)
+            probabilities = self.sklearn_model.predict_proba(input_data)[0]
+
+            # Get the predicted class and its probability
+            predicted_class = prediction[0]
+            class_probability = probabilities[predicted_class]
+
+            # Define class names for logging
+            class_names = ["ARTIST", "ALBUM", "SONG", "RANDOM"]
 
+            # Log the prediction details
+            self.logger.info(f"sklearn model prediction for '{artist_name}': "
+                            f"class={class_names[predicted_class]}, "
+                            f"probability={class_probability:.2f}")
 
+            # Check if the predicted class is "אמן" (0)
+            is_artist = predicted_class == 0
+
+            if is_artist:
+                self.logger.info(f"'{artist_name}' verified as an artist")
+            else:
+                self.logger.info(f"'{artist_name}' not verified as an artist. "
+                                f"Predicted as: {class_names[predicted_class]}")
+
+            return is_artist
+
+        except Exception as e:
+            self.logger.error(f"Error during sklearn prediction for '{artist_name}': {str(e)}")
+            return True  # Assume it's an artist if prediction fails
     def process_with_ner(self, text):
         if not self.nlp:
             return []
 
         doc = self.nlp(text)
-        return [ent.text for ent in doc.ents if ent.label_ == "SINGER"]
-
+        potential_artists = [ent.text for ent in doc.ents if ent.label_ == "SINGER"]
+
+        verified_artists = []
+        for artist in potential_artists:
+            self.logger.info(f"NER model identified potential artist: {artist}")
+            if self.verify_artist_with_sklearn(artist):
+                verified_artists.append(artist)
+                self.logger.info(f"sklearn model verified '{artist}' as an artist")
+            else:
+                self.logger.info(f"sklearn model rejected '{artist}' as an artist")
+
+        return verified_artists