AI for the intelligent automation of documents

Intelligent OCR thanks to Deep Learning AI in Konfuzio to capture documents faster and more accurately.

Provide structured information where your customers and users need it to automate processes or analyze data.

For what types of documents is Konfuzio suitable?

Invoices

A Invoice, cancellation invoice or credit note refers to any document that contains the calculation of a delivery or other service, regardless of how this document is referred to in business transactions. 

Read more about OCR with AI for invoices .

Payment advice

Payment notification contains a detailed list of all items that are settled by the payment. Details per item are the header data per invoice.

Read more about OCR with AI for payment notifications .

Insurance policies

An Insurance policy is a document confirming the successful conclusion of an insurance contract between the insurer and the insured. Insurance policies contain all important data of an insurance contract and are settled via premium invoices, see also Invoices.

Read more about possible applications for OCR with AI as insurer and pool

Vehicle documents

The Vehicle registration document is issued by the vehicle registration authority when registering or re-registering road vehicles and serves to identify a vehicle subject to registration.

Use Konfuzio for the OCR of vehicle registration documents.

Test free of charge

Get all the information in one email

    What are the options for Konfuzio IT operations?

    Server infrastructure

    Public Cloud (SaaS)

    You can rely on our highly secure, European data center with strict data protection requirements according to DSGVO. The public cloud is based on OpenStack, guaranteeing you maximum flexibility and future-proofing. You can start using Konfuzio immediately.

    Private Cloud

    Konfuzio in a private cloud is not provided to the general public, but only to selected users over the Internet or a private internal network. Private cloud computing extends the public cloud benefits with further customization options.

    On Premises Server

    On-Premises

    You operate Konfuzio under your own responsibility in your own data center. It is also possible that you operate Konfuzio on rented servers of a third-party data center. Probably the biggest advantage of on-premise software is in the area of data protection.

    What integrations does Konfuzio offer?

    API interface of Konfuzio

    REST API for online OCR with AI

    REST API is also referred to as machine-to-machine communication, since the various systems and devices are brought together and, in a sense, speak the "same language". Thanks to REST API, it is possible to distribute information and tasks among different servers and request them using an HTTP request.

    Power Automate flow

    RPA for intelligent automation of documents

    Robotic Process Automation (RPA) is an approach to process automation in which repetitive, manual, time-consuming or error-prone activities are performed by so-called software robots (bots) in a rule-based and automated manner. Bots use Konfuzio as a building block to recognize content in documents in the overall process.

    Documents are uploaded for processing with AI

    Web UI for the No-Code Training

    The business department streamlines processes via the browser-based web interface. Data is structured with a single click. Typing is a thing of the past. Exchange formats such as Excel or similar soften a uniform workflow. The UI is technically very easy to integrate into existing processes.

    Document AI Python SDK

    Document extraction with the Python SDK

    The SDK allows you to retrieve visual and textual features to create your own document models. Konfuzio Server serves as a user interface to define the data structure, manage training/testing data and expose your models as an API.

    For which languages is Konfuzio suitable?

    Afrikaans, Albanian, Asturian, Azerbaijani (Latin), Basque, Belarusian (Cyrillic), Belarusian (Latin), Bislama, Bosnian (Latin), Breton, Bulgarian, Buryat (Cyrillic), Catalan, Cebuano, Chamorro, Chinese (Simplified), Chinese (Traditional), Cornish, Corsican, Crimean Tatar (Latin), Croatian, Czech, Danish, Dutch, English, Erzya (Cyrillic), Estonian, Faroese, Fijian, Filipino, Finnish, French, Friulian, Gagauz (Latin), Galician, German, Gilbertese, Greenlandic, Haitian Creole, Hani, Hawaiian, Hmong Daw (Latin), Hungarian, Icelandic, Inari Sami, Indonesian, Interlingua, Inuktitut (Latin), Irish, Italian, Japanese, Javanese, K'iche', Kabuverdianu, Kachin (Latin), Kara-Kalpak (Latin), Kara-Kalpak (Cyrillic), Karachay-Balkar, Kashubian, Kazakh (Cyrillic), Kazakh (Latin), Khasi, Korean, Koryak, Kosraean, Kumyk (Cyrillic), Kurdish (Latin), Kirghiz (Cyrillic), Lakota, Latin, Lithuanian, Lower Sorbian, Lule Sami, Luxembourgish, Malay (Latin), Maltese, Manx, Maori, Mongolian (Cyrillic), Montenegrin (Cyrillic), Montenegrin (Latin), Neapolitan, Niuean, Nogay, Northern Sami (Latin), Norwegian, Occitan, Ossetian, Polish, Portuguese, Ripuarian, Romanian, Rhaeto-Romanic, Russian, Samoan (Latin), Scots, Scottish Gaelic, Serbian (Cyrillic), Serbian (Latin), Skolt Sami, Slovak, Slovenian, Southern Sami, Spanish, Swahili (Latin), Swedish, Tajik (Cyrillic), Tatar (Latin), Tetum, Tonga, Turkish, Turkmen (Latin), Tuvan, Upper Sorbian, Uzbek (Cyrillic), Uzbek (Latin), Volapük, Walser, Welsh, West Frisian, Yucatec Maya, Zhuang, Zulu.

    What file formats are supported?

    PDF, TIF, PNG, JPG, JPEG, EML, XLS, DOC, PPT are supported file formats

    PDF Extraction to structured information

    A reader of a PDF file should always be able to view and print the document in the form specified by the author. In PDF files, all information is stored as numbered objects. Objects include font information, character widths, character encodings used, page description, parameters for decoders, crop boxes, individual bookmarks, color definitions, page orders, bitmaps, forms, jump marks, and anything else that can be stored in PDF files. A one hundred page PDF file can easily contain 10,000 objects. Konfuzio finds the relevant information automatically.

    JPEG to text and JPEG to JSON

    The JPEG-Image format is supported by all operating systems. It is one of the most widely used image formats. The ratio of image quality to file size is good.

    JPG to text and JPG to JSON

    The JPG-Image format is supported by all operating systems. It is one of the most widely used image formats. The ratio of image quality to file size is good.

    PNG to text and PNG to JSON

    The PNG-The GIF format supports the alpha channel for transparency. This is newer and more modern than that of the GIF format. Moreover, compared to other image formats, the image format does not discard information when resaving.

    TIFF to text and TIFF to JSON

    No loss of quality is offered by the TIFF, as this is not compressed. High quality and accurate. In addition, the channel transparency is also supported again.

    EML Process files automatically

    Emails are internally divided into two parts: The header with headers and the body (text body) with the actual content of the message. In addition, further subdivisions are defined within the body. Furthermore, Konfuzio processes attachment, also called attachment, file attachment, or attachment, these are all files that are sent as an attachment to the text of an e-mail. Technically, this file is part of the body, but it is perceived as separate and treated so in common usage.

    MSG e-mails Process with attachment automatically

    Emails are internally divided into two parts: The header with headers and the body (text body) with the actual content of the message. In addition, further subdivisions are defined within the body. Furthermore, Konfuzio processes attachment, also called attachment, file attachment, or attachment, these are all files that are sent as an attachment to the text of an e-mail. Technically, this file is part of the body, but it is perceived as separate and treated so in common usage.

    Extract DOC and DOCX

    File formats for storing office documents, which are intended to enable data or file exchange between different office application packages. Due to the wide distribution of Microsoft Office, the associated binary and proprietary file formats of Microsoft Word, Microsoft Excel and Microsoft PowerPoint had established themselves as a de facto standard for document exchange in many areas. However, OpenOffice and Libre Office are also supported by Konfuzio.

    Extract PPT and PPTX documents

    File formats for storing office documents, which are intended to enable data or file exchange between different office application packages. Due to the wide distribution of Microsoft Office, the associated binary and proprietary file formats of Microsoft Word, Microsoft Excel and Microsoft PowerPoint had established themselves as a de facto standard for document exchange in many areas. However, OpenOffice and Libre Office are also supported by Konfuzio.

    Extract XLS and XLSX documents

    File formats for storing office documents, which are intended to enable data or file exchange between different office application packages. Due to the wide distribution of Microsoft Office, the associated binary and proprietary file formats of Microsoft Word, Microsoft Excel and Microsoft PowerPoint had established themselves as a de facto standard for document exchange in many areas. However, OpenOffice and Libre Office are also supported by Konfuzio.

    How does Konfuzio work?

    Our powerful machine learning solution extracts information quickly and accurately from any document to feed structured data into downstream systems.

    OCR for AI

    1. Segmentation per page

    Our Deep Computer Vision-based model has been trained with more than 100,000 documents and recognizes elements such as tables, paragraphs and headings across projects and regardless of language.

    add new annotation for training for AI

    2. text recognition by OCR

    Depending on the quality of the incoming document, we select the relevant technology for text recognition. Even on-premise, very good results can be achieved with an LSTM Deep Net.

    OCR text recognition of documents on text view

    3. Automatic reading

    Our Natural Language Processing Tool recognizes basic elements, so-called entities, for each language. Elements such as persons, places, companies and dates are combined with information from the computer vision model.

    Konfuzio SmartView

    4. Transfer learning

    The AI - the core of the Konfuzio AI - recognizes information in context. Thus, the street is assigned to the supplier, the invoice number to the invoice and the item number to an invoice item. In addition, the AI distinguishes between the address of the supplier and the recipient.

    API interface of Konfuzio

    5. API involvement & rules of the specialist department

    Before transferring to the target system, you can enrich the data from the API or CSV download. This gives you the option of applying further specialist rules if these are not already available in the target system. For example, you can check whether the VAT number is valid.

    Automate invoice receipt through AI and OCR

    6. Transfer to the target ECM system

    Our AI can be used immediately and yet continues to learn. Thus, the AI works in the background and the existing processes and interfaces can remain in place. The effort required to process documents is reduced to a minimum, and errors are avoided. To continuously improve the AI, you do not need an IT expert, but only training for the business department. Corrections made in the target system can be automatically reported to Konfuzio, which then learns automatically. 

    References

    Use cases for intelligent automation in data-driven enterprises

    Our customers use Konfuzio as a distribution center for unstructured data in documents, tweets, emails and many other texts. Through the AI platform, the right information is delivered directly to the responsible employees to enable better and faster decision making. The additional information in the respective documents provides additional added value for the data repository through its structuring and categorization and will not get lost for future, data-driven decisions.

    Arrow-up