embedding

embedding, data type

The embedding data type represents a dense vector derived from text. It is opaque and meant for similarity comparisons and arithmetic combinations. Embeddings can be written and read in .ion files, but they are not supported in .csv or .tsv files. Use spark to get a compact text representation. Embeddings support == and !=, as well as +, -, and * operators.

embedding(text: text) ?? embedding, function

def pure embedding(t: text): embedding

Returns the embedding vector associated with the input text.

Example

table T = with
  [| as Left, as Right |]
  [| "supply chain", "inventory optimization" |]
  [| "apple", "orange" |]

T.L = embedding(T.Left)
T.R = embedding(T.Right)
T.Score = similarity(T.L, T.R)
T.Label = spark(T.L)

show table "Embeddings" with
  T.Left
  T.Right
  T.Label
  T.Score

See also

User Contributed Notes
0 notes + add a note