Sogeti Cognitive Document Processing (CDP): 4 Week Implementation

Today, industries are highly regulated and require documentation, even in this digital age much of the paperwork is processed manually with an immense time and cost commitment. Manual processing of documents involves various steps of verification to remove errors and biases. All this is not just expensive and time consuming but also impacts the client experience.

Cognitive Document processing (CDP) is a cognitive Machine Learning based solution, that uses deep learning for classifying and extracting the relevant information from the unstructured documents. Leveraging Microsoft’s cognitive capabilities including artificial intelligence (AI) and machine learning (ML), it enables clients to ingest data from various sources, process and act on information extracted from unstructured documents. It can accelerate a wide range of document heavy processes like customer onboarding, claims processing, contract evaluation etc.

CDP leverages Microsoft’s various cognitive capabilities to process (classify the documents and extract relevant information) documents. Some of the key components used by CDP are Azure Cognitive services including computer vision, custom vision, form recognizer etc. And various Azure resources like App Services, Functions, registry services, PaaS components for deployment etc.

CDP is many times faster than manual processing and can work 24 X 7 with minimal oversight. It reduces the cost, effort, and risk by leveraging cognitive capabilities & frees up the time for the client resources to work on value producing tasks. CDP is offered both as an Cloud and as a service model. It can support documents in English, and in other languages such as Dutch, Swedish and other languages supported by Microsoft Computer vision.

CDP Business model includes below tasks: Discovery

1. Setup Work environment
2. Workshop with customer to understand the requirement and existing landscape
3. Setup Azure Services
4. Create the Project plan
5. Gather Data for training

Implementation

1. Deploy the CDP components to Azure
2. Train Model on Custom vision and Form Recognizer
3. Train custom models with CDP Framework for any custom object detection like stamps, sign etc.
4. Configure the models in CDP framework
5. Implement the API for the end to end pipeline
6. Model enhancements
7. QA and Bug Fixes

Deployment

1. User Acceptance testing
2. Deployment and training