Introduction

INFORMys is a flexible form reader system especially conceived for dealing with invoices and similar documents. Like most systems, INFORMys offers a twofold user interface for assisting the user during both form modeling and form reading. These different operations are supported by two modules, namely the Form Modeler (FM), which is charged for allowing the user to build the model (a graph) describing the forms of a specific class, and the Form Reader (FR), which implements the recognition engine. Fig. 1 shows the basic structure of the system.

  
Figure 1: The basic structure of INFORMys, a flexible form reader system, where the user can shape the layout of its own documents in the class of invoice-like documents.

The acceptable documents

The documents that are acceptable by the system contain information fields that cannot be located by relying on their absolute position, or by relying on their relationship with form lines. Some fields can be located only after having recognized the corresponding instruction fields. Typical examples of such forms are those delivered by companies for accounting their service (see Fig. 10).

With INFORMys we can deal with documents where four different kinds of ``objects'' can be found, namely lines, logos, instruction fields, and information fields. Our document class is identified by specifying all possible ways of relating the information to instruction fields and other objects. The class is defined specifying the objects and their relationships in the documents.

There are essentially four ways of locating an information field:

1. A field that is in a fixed position with respect to the layout lines. This field can be located after locating the corresponding reference lines.
2. An information field that is in a fixed position with respect to the corresponding preprinted instruction field (or with respect to a logo). This kind of field can be located by means of its absolute position in the form, or by the relative position with respect to the instruction field (or logo).
3. The instruction field (or logo) and the corresponding information field are not always in the same position in the form, but their mutual position is defined in advance. The information field can be located only after having recognized the related instruction field (or logo).
4. The information field is located by means of its position with respect to two other objects (e.g. two instruction fields) that are not always in the same position in the form. An example of this kind of fields comes from tables with instruction fields located in positions that are not known in advance.

The actual structure of the form arises, during the modeling phase, by properly combining instruction and information fields.

• The INFORMys Modeler
• INFORMys Reader