There are two possible ways to use Aleph Alpha’s semantic embeddings. If you have texts with a dissimilar structure (e.g. a Document and a Query) you would want to use asymmetric embeddings. Conversely, for texts with comparable structures, symmetric embeddings are the suggested approach.

​
Asymmetric

from langchain_community.embeddings import AlephAlphaAsymmetricSemanticEmbedding

document = "This is a content of the document"
query = "What is the content of the document?"

embeddings = AlephAlphaAsymmetricSemanticEmbedding(normalize=True, compress_to_size=128)

doc_result = embeddings.embed_documents([document])

query_result = embeddings.embed_query(query)

​
Symmetric

from langchain_community.embeddings import AlephAlphaSymmetricSemanticEmbedding

text = "This is a test text"

embeddings = AlephAlphaSymmetricSemanticEmbedding(normalize=True, compress_to_size=128)

doc_result = embeddings.embed_documents([text])

query_result = embeddings.embed_query(text)