Embeddings are numerical vector representations of text, images or other data. An embedding model translates a word, a sentence or a whole paragraph into a series of numbers with several hundred to several thousand values. This series of numbers does not map the letters, but the meaning.
The benefit lies in the geometry of this vector space. Terms with similar meanings end up close to each other, regardless of the exact choice of words. "Rechnung" and "Faktura" are assigned similar vectors, even though not a single letter matches. A computer can calculate this proximity and use it to measure similarity without anyone having previously maintained a dictionary of synonyms. Exactly this property makes embeddings the basis of semantic search: instead of searching for exact keywords, the system searches for closeness in content.
For SAP Business One- users will find this practically relevant as soon as knowledge from many sources needs to be searched. Manuals, support cases, blog articles and internal Documentation can be converted into embeddings and stored in a vector database. A user query is also translated into an embedding, the most similar text passages are retrieved and provided to a language model as context. This method has now become the standard way to supply a language model with up-to-date ERP knowledge without the need for costly retraining. Article master data and document texts can also be made searchable in this way, for example to track down duplicate suppliers or similar posting texts.
Demarcation
Embeddings are not a finished product, but an intermediate representation. The embedding model generates the vectors, the vector database stores and searches them, and semantic search is the use case that results from this. Anyone who talks about "embeddings" is therefore referring to the numerical representation itself, not the search interface and not the model that calculates them.
AI in everyday work: Here's how to find the right start
AI assistant meets SAP Business One: Two paths out of the Versino AI shop
Sprint and Swirl: How an AI agent divides work and learns from trials
LATS: Considering several paths simultaneously with AI — and the availability check in SAP B1
ReAct and RLEF: How Artificial Intelligence Learns from Real Mistakes — and What Automated Bank Reconciliation in SAP Business One Has to Do with It
