Intelligent Document Processing

In many companies, incoming mail is scanned and stored in the archive exclusively as image files. It is not possible to search automatically in these image files, neither within a document nor across many documents. A remedy for this is the text layer, which is saved together with the image file, e.g. as a PDF.

• The text layer is the basis for full-text search in complete archives, but also for copying text parts or intelligent, automated further processing in other applications.

Even though handwritten documents are on the decline, many companies still have a considerable amount of handwritten free text (as incoming mail from end customers). In addition, there are handwritten notes, file numbers and incoming stamps on many forms, which are not or incorrectly recognised by most OCR systems.

• TIMETOACT GROUP offers a solution module that has very good handwriting recognition and also transcribes these contents into the text layer. In this way, handwritten input documents can also be fed into automatic further processing. These documents are also made fully searchable automatically.

Recognition errors and blurring occur with every scanning and OCR process. Individual characters or word components are not recognised correctly, for example. These errors remain or have to be corrected manually by humans.

• The PerceptionMatrix (©Planet AI GmbH) provides a remedy. In this module, not only individual characters but the larger context is integrated. In this way, errors can be partially corrected automatically or various recognition alternatives with corresponding probability values (scores) are automatically displayed. In addition, the corresponding text passages can be called up and displayed for post-processing. The PerceptionMatrix thus increases the recognition quality and is the basis for tool-supported, efficient post-correction, error-tolerant search or further processing.