Pre-Printed Form Recognition and Extraction of Data

Forms are one of the most common classes of documents which organizations encounter and official communication through pre-designed forms is now a common practice. These forms provide a space for entering the transaction details and the problem of form understanding is recognizing the form followed by extracting the information contained in each of its variant fields. Since forms are meant for very specific applications, one usually expects very high accuracy and processing speeds. The problem of information processing from forms is more structured in nature, while the recognition part is complex since the texts are usually handwritten. Document processing is an important step in office automation and in this paper a method is proposed in which form type is recognized in terms of its heading and title layout composition to match the process of recognition as closely as possible to the human understanding system. For this purpose knowledge of different forms used in an organization is created by suitable apriori learning. Extraction of the information requires the recognizing the entries made in the space provided in the forms. Performance and evaluation of the method has shown promising results.