Automatic content analysis that identifies entities, relationships, risks and relevant information.
The information in documents is not in the metadata
A document system usually knows who sent a document, when it was registered and which folder it sits in. It does not know what the document says. And decisions are made on what it says.
The consequence: for any question about content — which contracts expire next quarter, which penalty clauses we have, where a given party appears — someone has to read.
What content analysis extracts
Entities: people, organisations, addresses, amounts, dates, identification numbers. They become fields you can search and filter on, not text buried in the document.
The relationships between them: who is a party to what, which document refers to which other, which obligation falls to whom. These links feed the organisation’s knowledge graph.
Elements of potential risk: approaching deadlines, clauses unusual for the organisation’s pattern, values above set thresholds. The system flags them; the assessment stays with people.
The information relevant to the workflow the document enters — what must be filled in, what is missing, what must be checked before approval.
- entities turned into searchable fields;
- relationships between documents, people and organisations;
- flagging of deadlines and unusual elements;
- automatic filling of the data for the next workflow step.
What the organisation gains
Documents become queryable. Questions about the contract portfolio, exposures or deadlines no longer need a manual review, but a query.
And at the operational level, the clerk receives the document with the data already extracted and what must be checked already marked, instead of transcribing and searching.
Exact-word search misses exactly what you are looking for
Someone looking for “the contract with the company that services our lifts” knows neither what the document is called nor which words it contains. A classic exact-match search returns either nothing or three hundred results.
The problem grows with the age of the archive. Documents written ten years ago use other terms than last year’s, and today’s searcher has no way of knowing them.
Search by meaning, not by letter
Semantic search compares the meaning of the question with the meaning of the documents, not character strings. In practice: a query in plain language finds the document even when it uses other words for the same thing.
Search covers both the content of the documents and their metadata — parties, dates, registers, deadlines — so questions that mix the two kinds of criteria are answered by a single query.
Results are filtered by permissions. A user sees only the documents they have access to; search does not become a back door for finding out what exists without the right to see it.
- queries in plain language, not in search syntax;
- covers both the content and the metadata of documents;
- results filtered by the user’s permissions;
- works on old documents too, with other terminology.
What the organisation gains
Time lost finding documents is one of the biggest hidden wastes in an organisation, precisely because it never shows up as a separate activity. It only shows as general slowness.
And at an inspection or a request with a deadline, the difference between finding a document in two minutes and finding it in two days changes the nature of the answer.
How to work
ECM — documents and registry
VOGO ECM is the platform for the life cycle of an institution’s documents: electronic registry, document management, cases and workflows, archive, security and compliance. For day-to-day work, see the articles on the Explorer, folders and permissions, document properties, registering in a register, export, the registration receipt and incoming e-mails.
The document Explorer
The Explorer has three panels: the tree on the left, the list of documents in the middle, and on the right the card of the selected document, which can be collapsed. The View selector above the tree changes how documents are grouped: by folders (the default) or by the classification structure. The list shows the standard columns and the fields of the folder’s document type, as many as fit on the screen.
New folder and uploading documents
In the Explorer, choose the folder, then:
- Folder nou (New folder) creates a subfolder; it receives the permissions of the parent folder. - Încarcă (Upload) adds files: each file becomes a document of the chosen type, among the types the folder accepts, and receives the folder’s permissions.
Both need the Scriere (Write) permission on the folder. Top-level folders are created only by the ECM administrator.
Renaming and deleting a folder
Renaming and deleting a folder need the Full permission on it. Only an empty folder can be deleted: move or delete its documents and subfolders first.
Permissions on folders and documents
Permissions work as in Windows. You change them in the Securitate (Security) section of the folder card, in the Explorer. The levels:
- Doar previzualizare (Preview only): the document can be seen, not downloaded; - Citire (Read): you see and download; - Scriere (Write): you also upload to the folder; - Full: you also change permissions, rename, delete.
Interzis (Denied) beats Permis (Allowed). A change applies to the folder only, or also to all its subfolders and files, as you choose. A new subfolder or document takes the permissions of the folder it is created in.
The properties of a document
The properties of a document — name, type, the fields of the type, access level — are shown and changed in the Proprietăți (Properties) tab. To change them you need the edit permission on the document.
Starred documents and documents shared with me
In the Explorer, the Cu stea (Starred) view shows the documents you have starred; the star is yours, other users do not see it. The Partajate cu mine (Shared with me) view shows the documents someone else has given rights to, to you or to a group you belong to.
Questions and answers
Is every document analysed?
It is configured by category. Not every document type justifies the same depth of analysis.
Does the extracted data replace human checking?
No. They are proposals, and the level of confirmation needed is set per type of operation.
What happens with documents the analysis gets wrong?
The operator’s correction stays on record and shows, over time, where the rules need adjusting.
Does the analysis see documents the user has no access to?
Analysis results follow the same permissions as the source documents.
Does it search scanned documents too?
Yes, to the extent their content was extracted at capture. Scanned documents without text extraction remain findable by metadata.
Do users see documents they have no access to?
No. Results are filtered by permissions before they are shown.
Do we have to learn a search syntax?
No. The question is asked in plain language. Exact criteria remain available for those who prefer them.
How fast do new documents appear in search?
Indexing is done when the document enters the system. The actual delay depends on volume and is sized at implementation.