Responses are received by
• Mail • Fax • Emails • Web • SOAP Message

• Incoming Mail/Scanning
Responses are received as input documents. They are scanned and appear in the Papyrus Scan Client with thumbnails of all the scanned documents (TIFF Images) and the first one displayed in full.

• Incoming Faxes
Fax documents are visible in the Input Queue displaying the originating fax number.

• Incoming Emails
For emails, neural networks are used to compare similarities and differences with incoming emails and those stored in the domain knowledge - built up by examples. Words and phrases are used to classify an email, and not just those in the subject line. This is approximately 10 times quicker then OCR recognition.

• Web Response
You can have a Web application where customers fill out HTML response forms. These are captured with the HTTP Adapter.

• SOAP Message:
You can use a SOAP message to send response information from your 3rd party application server, JAVA application, Web Portal or any other WebServices enabled system. The SOAP message triggers a response task.

1

2

3

4

5

Responses are received as input documents: mail, fax, Web, email, SOAP message

Classification of responses is automatically performed. Unrecognized documents are displayed for visual classification in Papyrus WebClassify. Self learning technology ensures that all types of documents can be classified.

Content Extraction
All fields of interest are extracted from the recognized document class.

The agent performs rule driven processing of all communication tasks.

A response is created. Following options apply:

Self-learning technology ensures that all types of documents can be classified.

Classification of responses

A physical document, fax or email is received. The classification is independent of document type. The classification is trained by example.

Exception handling:
If the classification is not successful, then it is routed to the manual classification process. Each unrecognized document will be displayed automatically for visual classification. The document will be flagged and routed to the supervisors inbox for further verification.

The classification applies the knowledge of rules acquired in training unknown documents. The system uses a number of documents for each document-class for training purposes. From this input the classification process learns the rules by itself.

Fine-tuning and retraining the system:
Based on the documents that could not be classified correctly, constant fine-tuning is possible while running in real time mode. This ensures long-term optimization of the system and enables flexible adjustment of the rules based on the changes in the documents that are received.

Manual Classification

Web Classify
All unrecognized documents that failed to be classified to a known document type are still in the Extraction Queue. These need to be classified manually. Manual classification is performed automatically using an HTML page that allows an operator to manually classify each unrecognized document in sequence against a pre-defined set of document types.

Each manual classification will be stored against the logical document Template, and when the next ‘training’ process is run, these manual classifications will be added to the domain knowledge. Any subsequent document will be classified automatically, not requiring manual classification.