cosine similarity

main.py

@@ -1,19 +1,38 @@
+import warnings
 from transformers import AutoTokenizer, AutoModel
 import torch
+import torch.nn.functional as F
+
+# Suppress FutureWarnings and other warnings
+warnings.simplefilter(action='ignore', category=FutureWarning)
 
 # Load the tokenizer and the model
 tokenizer = AutoTokenizer.from_pretrained('allenai/scibert_scivocab_uncased')
 model = AutoModel.from_pretrained('allenai/scibert_scivocab_uncased')
 
-# Prepare a test input sentence (e.g., "Hello, world!")
-input_text = "Hello, world!"
-
-# Tokenize the input text and convert it to input IDs
-inputs = tokenizer(input_text, return_tensors="pt")  # Return tensors in PyTorch format
-
-# Forward pass through the model
-with torch.no_grad():  # Disable gradient calculation since we are only doing inference
+# Function to compute sentence embeddings by pooling token embeddings (CLS token)
+def get_sentence_embedding(text):
+    inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True, max_length=512)
+    with torch.no_grad():
         outputs = model(**inputs)
     
-# Output model's hidden states (for the last layer)
-print(outputs.last_hidden_state)
+    # Pooling strategy: Use the hidden state of the [CLS] token as the sentence embedding
+    cls_embedding = outputs.last_hidden_state[:, 0, :]  # Shape: (batch_size, hidden_size)
+    return cls_embedding
+
+# Example subject and abstract
+subject = "Artificial Intelligence in Healthcare"
+abstract = """
+Artificial intelligence (AI) is transforming healthcare with its ability to analyze complex medical data and assist in diagnosis. 
+AI models, especially in medical imaging, have shown promise in detecting diseases like cancer and predicting patient outcomes.
+"""
+
+# Get embeddings
+subject_embedding = get_sentence_embedding(subject)
+abstract_embedding = get_sentence_embedding(abstract)
+
+# 2. **Measure Semantic Similarity Using Cosine Similarity**
+
+# Compute cosine similarity between subject and abstract embeddings
+similarity = F.cosine_similarity(subject_embedding, abstract_embedding)
+print(f"Cosine Similarity: {similarity.item():.4f}")