Embeddings are a way to transform texts, documents, or individual words into numbers—more precisely, into mathematical vectors
In this way, documents (or parts of them) are represented within a continuous vector space. This transformation enables the software to understand the meaning of the content, not just the individual words it contains.
In practice, thanks to embeddings, two texts with a similar meaning are positioned close to each other, while texts with different meanings are placed farther apart.
As a result, documents related the same domain—even if written in different ways—are recognized as conceptually similar.
Embeddings are important because traditional computer systems work well with numbers, exact word matches, and precise search rules, but struggle with natural language and ambiguity.
Embeddings bridge this gap by allowing software to interpret human language, understand the context of a search, and reason based on semantic similarity rather than exact matches. This makes it possible to efficiently identify similarities between documents and to implement advanced features such as semantic search.
The system understands what you are looking for, not just what you type.
Thanks to this capability, the system can analyze document content, identify relationships, suggest relevant attachments, and automatically link similar documents. For example, it can recognize a contract as being similar to other agreements already stored in the archive.
The process of calculating a document’s embedding is not unique and depends on the model used. LogicalDOC supports the definition of multiple embedding schemes, based on specific embedding models that can be implemented either directly within the system or through external services.
Benefits of this feature
- Advanced semantic search
- Automatic document classification
- Linking between related documents
- Easily usable even for non-technical users: access to information becomes simple and intuitive for everyone.
