Platform (Generation 3)

intelligent document processing

Our Intelligent Document Processing (IDP) platform has grown from 5 years of practical challenges in efficiently and reliably extracting information from complex document collections and streams for our customers. Humans are excellent at recognizing appropriate patterns, but as volume and complexity increase, this becomes a laborious and unreliable routine task.

Artificial intelligence, especially machine learning, has initiated a paradigm shift in information technology: until now, (human) intelligence had to recognize patterns, rules and heuristics and translate each individual case into program code. Now the machine can learn these itself using examples.

What makes our solution different

We don't just talk about AI, we have integrated it into the inserve IDP platform consistently, with a sense of proportion and a focus on practical added value:

Training instead of programming

Speed and adaptability through consistent use of AI building blocks.

Efficient training

Specialized AI models that are as simple as possible plus Active Learning minimize the need for training.

Z

Complete processing chain

Coverage of all necessary steps (structuring, extraction, analytics).

High detection rates

High performance of AI models through clever use of all information (text, structure, image).

Convenient use

Ergonomic tools for all tasks - flexibly configurable, but as simple as possible.

Modern technology

Scalable and reliable operation - flexible modes of operation (on premise, public/private cloud).

Processes and building blocks

The core processes of information extraction and predictive / prescriptive analysis are embedded in an overall solution with all the necessary building blocks for collaboration, data exchange and control of all processes:

Interaction between AI and user

The IDP platform combines the Human-in-the-Loop principle with Active Learning (automatic selection of unsafe recognized examples) to enable the user (trainer) to specifically and efficiently achieve high recognition performance with as few training examples as possible.

Complete processing chain for information extraction

The IDP platform covers all necessary processing steps in a flexibly configurable and highly automatable way:

 

  • Integration of different data sources via various interfaces
  • Reliable Text recognition and subsequent recognition of words
  • Separation of incoming batches into individual documents
  • Structuring, clustering, classification of stacks, documents, pages
  • Extraction desired information
  • Summary of information on entities
  • Validation of completeness and consistency
  • Provision ofdocuments and information for subsequent processes

Combinable processing steps & workflows

The modularly designed, functional as well as AI-based processing steps can be configured into powerful and intelligent workflows as required.

By using and combining efficient machine learning methods from the fields of machine vision, natural language processing, predictive/prescriptive analytics and active learning that are specialized for the individual tasks, the IDP platform - automated in the background - achieves a high training efficiency and recognition performance for the users.

Automatic structuring: Clustering

We use the content structure and the spatial arrangement - commonly referred to as the "layout" - of information in documents to organise large and heterogeneous document collections automatically, i.e. without training.

This not only facilitates the initial access, but also significantly accelerates and improves the subsequent processing steps for users and AI. Analogous to human pattern recognition, the recognized clusters are used, for example, to support document classification and information extraction.

Predictive/Prescriptive Analytics:

Forecast, evaluation and recommendation

However, even more can be done from the information about structures and contents obtained across documents: by applying predictive and prescriptive analytics models to the structured, possibly temporally ordered data, we obtain further information:

U

Anomaly detection

Unexpected deviations or trend breaks provide indications of transactions to be reviewed

b

Estimation and prediction

Closing gaps in document inventories and forecasting future developments

l

Evaluation and recommendation

Assessment of risks, customer satisfaction and recommendation or automation of decisions

Modern technology

By leveraging and seamlessly integrating modern open source technologies at an industrial level, the IDP platform is cost-efficient and flexible to fit your circumstances, combining the worlds of robust enterprise software and innovative AI solutions - future-proofing you for the future:

Flexible run

The trend towards the public cloud is unstoppable, but in many situations this is not desired, e.g. for reasons of data protection. Operation in the private cloud or in your own data center also offers individual advantages. The IDP platform can be used flexibly in each of these variants.

Scale up

Efficient use of available resources as well as parallelization and distribution enable high-performance processing for all sizes of processing loads and high scalability in the event of load fluctuations or peaks.

Seamless monitoring

Our monitoring and alerting tools not only ensure that you achieve high operational reliability by monitoring all technical parameters. You also keep an eye on the robust processing and detection performance of the AI modules.

Use services

Depending on your requirements and circumstances, you can also use the IDP platform flexibly through our service options: in full service, in managed service or independently in software-as-a-service mode.

We are at your disposal for any questions you may have