add omartificial to app.py

app.py

@@ -8,6 +8,7 @@ import numpy as np
 # Load models
 model = SentenceTransformer("distilbert-base-multilingual-cased")
 modela = SentenceTransformer("paraphrase-multilingual-MiniLM-L12-v2")
+modelb = SentenceTransformer("Omartificial-Intelligence-Space/Arabert-all-nli-triplet-Matryoshka")
 
 # Load data
 df = pd.read_csv("cleaned1.csv")
@@ -19,10 +20,16 @@ embeddings = torch.load("embeddings1_1.pt")
 embeddings2 = torch.load("embeddings2_1.pt")
 embeddings3 = torch.load("embeddings3_1.pt")
 
+
 embeddingsa = torch.load("embeddings1.pt")
 embeddingsa2 = torch.load("embeddings2.pt")
 embeddingsa3 = torch.load("embeddings3.pt")
 
+
+embeddingsb = torch.load("embeddingso1_3.pt")
+embeddingsb2 = torch.load("embeddingso2_3.pt")
+embeddingsb3 = torch.load("embeddingso3_3.pt")
+
 # Extract questions and links
 df_questions = df["question"].values
 df_links = df["link"].values
@@ -127,14 +134,17 @@ def predict(text):
     # Semantic similarity scores
     query_embedding = model.encode(text, convert_to_tensor=True)
     query_embeddinga = modela.encode(text, convert_to_tensor=True)
-
+    query_embeddingb = modelb.encode(text, convert_to_tensor=True)
     # Cosine similarities (averaged from two models)
     sim_scores1 = (util.pytorch_cos_sim(query_embedding, embeddings)[0] + 
-                   util.pytorch_cos_sim(query_embeddinga, embeddingsa)[0]) / 2
+                   util.pytorch_cos_sim(query_embeddinga, embeddingsa)[0] +
+                   util.pytorch_cos_sim(query_embeddingb, embeddingsb)[0] ) / 3
     sim_scores2 = (util.pytorch_cos_sim(query_embedding, embeddings2)[0] + 
-                   util.pytorch_cos_sim(query_embeddinga, embeddingsa2)[0]) / 2
+                   util.pytorch_cos_sim(query_embeddinga, embeddingsa2)[0] + 
+                   util.pytorch_cos_sim(query_embeddingb, embeddingsb2)[0])/ 3
     sim_scores3 = (util.pytorch_cos_sim(query_embedding, embeddings3)[0] + 
-                   util.pytorch_cos_sim(query_embeddinga, embeddingsa3)[0]) / 2
+                   util.pytorch_cos_sim(query_embeddinga, embeddingsa3)[0]+ 
+                   util.pytorch_cos_sim(query_embeddingb, embeddingsb3)[0]) / 3
 
     # BM25 scores
     bm25_scores1 = compute_bm25_scores(text, bm25_model1,corpus_length1)