EMBEDDINGS-EDI

EDI text embeddings, AI & data integration applied to EDI.

Definition

EDI embeddings are vectors (typically 768 or 1536 dimensions) computed by an embedding model (OpenAI text-embedding-3, Cohere embed v3, BGE) over fragments of TDED, EANCOM manuals, X12 examples. Enable semantic search beyond keyword match — "tax" retrieves TAX, MOA+124, BT-118, etc.

Origin

Concept word2vec (Mikolov 2013), GloVe (Pennington 2014), generalised with BERT then OpenAI Ada (2022).

Example in context

Index for 30,000 fragments of EDI manuals, queries via cosine similarity over OpenAI text-embedding-3-small.

Related terms

• RAG EDI — main use case.