Document classification means sorting documents in a way that makes it easier to locate them later. For example, you classify a document as a sales order, as an order from a particular client, for a particular product and of a particular date. With this information, you can retrieve a particular order, all orders from a particular client, or for a certain product, and so on.
1. It is how users tend to look for a document that determines how it will be classified. In the example of the sale order above, the different classification criteria are all ways users tend to ask for order details. They might want to review a particular order, or all orders from a client, or all orders for a product and so on.
2. Structured documents such as sales orders are stored in databases with defined structures. Database queries can then retrieve them by desired criteria and generate reports providing desired information.
3. Unstructured data such as correspondence, e-mails, reports, etc. cannot be so easily stored in structured databases. Instead, they tend to be indexed by document metadata that contain brief information about the topic covered. For example, you might want to retrieve all reports on market conditions for a specific product.
4. Indexing by metadata will really work only if there are some standards for attaching the metadata. It must contain standard information, such as date of creation, author, and topic covered by the document. Secondly, similar documents must be described similarly by all persons. To achieve this, choice lists are typically standardized and users are provided drop-down selection boxes to select one of these standard choices.
5. Metadata can be extracted automatically by the system when a document is created, such as the date of creation, or entered manually by the user, as for the topic selection.
6. Full text search enables documents to be selected by words in the document content. However, this is likely to provide unsatisfactory results as the same words might occur in many documents and the search will result in too many documents.
7. One solution to having too many search results is to combine a hierarchical directory structure with search capabilities. Documents are stored in directories and subdirectories with meaningful names, and you browse to the relevant subdirectory before invoking a search command limited to that directory.
8. Classification and tagging of documents can serve purposes other than retrieval. For example, meta-tagging documents with their retire-by dates can help programs to retrieve all documents that have expired and even dispose them as instructed by another meta-tag. This can reduce storage media costs by freeing storage space.
9. Documents can also be tagged by their business-sensitivity. Documents tagged as highly sensitive can then be made accessible subject to specific restrictions applied automatically.
10. Document classification can thus serve multiple objectives. A Microsoft blog (http://blogs.technet.com/filecab/archive/2009/05/11/windows-server-2008-r2-file-classification-infrastructure-managing-data-based-on-business-value.aspx) reports that the most frequent tagging requirements are Personal Information (yes/no), Business Criticality, Confidentiality, Project, and Retention Period. If documents are assigned properties accordingly, systems can automate several document-related tasks leading to the kinds of business benefits mentioned in the blog.
Document classification cannot be an ad-hoc exercise carried out by the document creators. Instead, it must follow standard conventions that have been developed with specific attention to desired objectives. These objectives can include retrieval, retention and confidentiality objectives.