What is semantic similarity in AI
gpt-4o or claude-3-5-sonnet-20241022 to quantify meaning similarity for tasks like search and clustering.How it works
Semantic similarity works by converting text or data into numerical vectors that capture their meaning in a high-dimensional space. Imagine each sentence as a point in a multi-dimensional map where closer points mean more similar meanings. Models like gpt-4o generate these embeddings by analyzing context, synonyms, and relationships beyond exact words, enabling AI to understand that "car" and "automobile" are closely related.
Concrete example
Here is a Python example using the OpenAI SDK to compute semantic similarity between two sentences by comparing their embeddings:
import os
from openai import OpenAI
client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])
sentences = [
"I love playing football.",
"Soccer is my favorite sport."
]
# Get embeddings for each sentence
embeddings = []
for sentence in sentences:
response = client.embeddings.create(
model="text-embedding-3-large",
input=sentence
)
embeddings.append(response.data[0].embedding)
# Compute cosine similarity
def cosine_similarity(vec1, vec2):
import numpy as np
vec1 = np.array(vec1)
vec2 = np.array(vec2)
return np.dot(vec1, vec2) / (np.linalg.norm(vec1) * np.linalg.norm(vec2))
similarity_score = cosine_similarity(embeddings[0], embeddings[1])
print(f"Semantic similarity score: {similarity_score:.4f}")Semantic similarity score: 0.8723
When to use it
Use semantic similarity when you need to compare meaning rather than exact text matches. It is essential for search engines, recommendation systems, clustering documents, detecting paraphrases, and question-answering systems. Avoid it when exact string matching or syntactic analysis is required, such as code syntax checking or strict data validation.
Key terms
| Term | Definition |
|---|---|
| Semantic similarity | A metric that quantifies how much two texts share meaning. |
| Embedding | A numerical vector representing text or data in a high-dimensional space. |
| Cosine similarity | A measure of similarity between two vectors based on the cosine of the angle between them. |
| Vector space | A mathematical space where embeddings are represented as points or vectors. |
Key Takeaways
- Semantic similarity captures meaning beyond exact words using vector embeddings.
- Use semantic similarity for search, recommendations, and paraphrase detection.
- Cosine similarity is a common method to quantify semantic similarity between embeddings.