Upload 4 files

app.py

@@ -0,0 +1,67 @@
+
+import gradio as gr
+import joblib
+import json
+import numpy as np
+import pandas as pd
+
+from sklearn.metrics.pairwise import cosine_similarity
+from sentence_transformers import SentenceTransformer
+
+# טוען את המודל והקבצים
+gmm = joblib.load("gmm_model.pkl")
+
+with open("cluster_to_emotion.json", "r") as f:
+    cluster_to_emotion = json.load(f)
+
+# טוען את מאגר השירים
+song_db = pd.read_parquet("hf://datasets/johanf/taylor-swift/data/train-00000-of-00001.parquet")
+song_db = song_db[["lyrics", "title"]].dropna().drop_duplicates()
+song_db["lyrics"] = song_db["lyrics"].str.strip()
+song_db["title"] = song_db["title"].str.strip()
+song_db = song_db.reset_index(drop=True)
+
+# מחשב embedding לכל השירים
+embedding_model = SentenceTransformer("all-MiniLM-L6-v2")
+lyrics_list = song_db["lyrics"].tolist()
+lyrics_embeddings = embedding_model.encode(lyrics_list, show_progress_bar=True)
+
+# מודל להמרת טקסט לרגש
+emotion_model = SentenceTransformer("j-hartmann/emotion-english-distilroberta-base")
+
+def predict_emotion(text):
+    embedding = emotion_model.encode([text])
+    cluster = gmm.predict(embedding)[0]
+    return cluster_to_emotion[str(cluster)]
+
+def find_matching_song_by_emotion(user_input):
+    emotion = predict_emotion(user_input)
+
+    # מוצא שירים שמתאימים לרגש הזה
+    candidates = song_db[song_db["lyrics"].str.lower().str.contains(emotion.lower())]
+
+    if candidates.empty:
+        candidates = song_db
+
+    user_embedding = embedding_model.encode([user_input])
+    candidate_lyrics = candidates["lyrics"].tolist()
+    candidate_embeddings = embedding_model.encode(candidate_lyrics)
+
+    similarities = cosine_similarity(user_embedding, candidate_embeddings)[0]
+    top_idx = np.argmax(similarities)
+    
+    title = candidates.iloc[top_idx]["title"]
+    lyrics_snippet = candidates.iloc[top_idx]["lyrics"][:200].replace("\n", " ")
+    score = similarities[top_idx]
+
+    return f"**{title}**  (match: {score:.2f})\n\n`{lyrics_snippet}...`\n\n_Emotion: {emotion}_"
+
+demo = gr.Interface(
+    fn=find_matching_song_by_emotion,
+    inputs=gr.Textbox(placeholder="Tell me something that happened today"),
+    outputs="markdown",
+    title="Taylor Swift Mood Matcher",
+    description="Tell me what you're feeling and I’ll match you with a Taylor Swift song that fits your mood."
+)
+
+demo.launch()

cluster_to_emotion.json

@@ -0,0 +1 @@
+{"0": "sadness", "1": "joy", "2": "fear", "3": "anger", "4": "sadness", "5": "surprise"}

gmm_model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b53965ce9db017583febb2c1461da646f37d1f81066ca147b090d9511707b72
+size 84972567

requirements.txt

@@ -0,0 +1,6 @@
+gradio
+pandas
+numpy
+joblib
+scikit-learn
+sentence-transformers