Today we’re launching “Parashift Document Center”, Parashift’s solution for business process outsourcers and large enterprises.
The existing range of extraction APIs will be replaced by a comprehensive product suite that reflects the learning we have made with customers over the past 6 months. Below is an overview of the current and future features of “Parashift Document Center” (PDC).
Three different processing options are now available to our customers:
With the option “Basic-Extraction” all fields of a document defined by Parashift are read from the document and delivered within 10 to 40 seconds via the Parashift API. This option is ideal for software vendors who want to use e.g. invoice information like IBAN, date or total amount in their application. (The Basic Extraction can be tested directly here)
The “Self-Validation” option additionally provides the customer with an interface for the post-validation of data. The Parashift Extraction reads all data in the first step (Basic-Extraction) and makes it available in a workflow for post-processing. Where necessary, users can validate and automatically deliver the data via the Parashift API.
The option “Full-Extraction” completely saves Parashift clients from having to validate the results. Fully validated extraction results are delivered via the Parashift API. With this processing method, customers can dispense with all manual work in connection with data extraction. With its own QA team and various technological measures, Parashift ensures the appropriate quality.
The processing option can be decided flexibly by parametrizing the call on the Parashift API. This means that only one API must be integrated, which saves integration cost.
Ongoing radical improvements in accuracy
In the past 24 months, Parashift has achieved a high extraction accuracy for the document type invoice for Switzerland and Germany. From different benchmarks of our customers Parashift emerged as the provider with the highest accuracy. While the achieved accuracy is already great, even our own team is amazed at the rate of improvement. We would have expected the weekly improvement to gradually level off after 24 months. The opposite is the case; the improvement rate even increases over time.
We are therefore confident that we will make great strides towards our goal of being able to extract all business documents of the same quality as a human being in less than 10 seconds at a cost of less than 1 ct. We still have a lot of work to do, but the thesis we put forward at the beginning, that there is a technologically and economically valid path to achieve this goal, is confirmed more and more.
“Document Network” – everyone learns from everyone
The reason for these improvements, besides the proprietary machine learning cluster of Parashift, is that our models learn from all processed documents. We call this part of our infrastructure the “Document Network”. While the actual data of the documents is strictly protected and separated, the learnings gained from the processing are quickly integrated into the machine learning models and are thus available to all customers from day 1. We plan to open the “Document Network” on the validation page for customers later, so that validation teams can fill their idle time with work from other Parashift platform participants, which significantly helps to further reduce the cost of extraction. We will also open the network to external part-time/home workers.
While Parashift Document Center is currently trained on the document type invoice, we will continue to expand in the coming months:
A wide range of standardized document types
Together with interested parties and customers, we have compiled a list of around 50 document types, all of which we will implement over time as so-called Parashift standard document types. These include other accounting document types as well as document types for logistics, credit processing, real estate management and private documents. For each document type we define the standard fields, which are necessary for further processing in 80% of all use cases. With our technology, we train them to the same level as today’s invoices and then continuously improve their accuracy.
Customer-specific, individual document types
In addition to the standard document types, you will also be able to create individual document types in Q4/2019 (tentative). This means that any type of document type can be taught-in. In addition, a standardized Parashift document type can be used as a template. For example, it is possible to extract an invoice and to train and extract completely customer-specific fields. This allows private, individual learning to be combined with the fleet learning of the “Document Network”.
Expansion of the Validation Interface
In the coming months, we will significantly expand the Validation Interface to cover various scenarios and document types.
It is expected that in Q3 we will also launch a processing option for document classification only. This will allow documents to be classified quickly and the subsequent processing mode to be determined.
In recent months, we have invested significant effort in compliance with the DSGVO. We can now offer processing in all compliance zones that is fully DSGVO compliant. The Parashift specific compliance zones include “Switzerland”, “EU” and “Worldwide” and relate to data storage and processing.
Access and pricing
The pricing of Parashift Document Center is super simple: It consists of a monthly subscription of EUR 549 (2,000 documents per month already included) and a processing price per document that is graded according to volume. The subscription includes application support and operation. There are no license or annual maintenance costs. Any processing option can be combined.
Getting started with the product is very easy: no demos and lengthy setups are necessary. Simply open your Parashift Document Center account and start the 14-day trial immediately. We will also be happy to advise you.