File size: 1,998 Bytes
d6c005e | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 | import numpy as np
from itertools import combinations
from sentence_transformers import SentenceTransformer
embedding_model = None
def load_embedding_model():
global embedding_model
if embedding_model is not None:
return
print("Loading sentence transformer model...")
embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
print("Embedding model loaded.")
# ── Cosine Similarity ──────────────────────────────────────────────────────
def cosine_similarity(vec_a: np.ndarray, vec_b: np.ndarray) -> float:
dot_product = np.dot(vec_a, vec_b)
magnitude = np.linalg.norm(vec_a) * np.linalg.norm(vec_b)
if magnitude == 0:
return 0.0
score = dot_product / magnitude
return float(np.clip(score, 0.0, 1.0))
# ── Main Entry Point ───────────────────────────────────────────────────────
async def compute_similarity(texts: dict) -> list[dict]:
load_embedding_model()
valid_texts = {
strategy: text
for strategy, text in texts.items()
if text and isinstance(text, str) and text.strip()
}
if len(valid_texts) < 2:
return []
strategies = list(valid_texts.keys())
text_list = list(valid_texts.values())
print(f"Embedding {len(text_list)} texts...")
embeddings = embedding_model.encode(text_list, convert_to_numpy=True)
print("Embeddings computed.")
scores = []
for i, j in combinations(range(len(strategies)), 2):
score = cosine_similarity(embeddings[i], embeddings[j])
scores.append({
"strategy_a": strategies[i],
"strategy_b": strategies[j],
"score": round(score, 4),
})
scores.sort(key=lambda x: x["score"], reverse=True)
return scores |