Empowering Intelligent Work through Infrrd’s Improved IDP Platform
In keeping with our spirit of continuous learning and improvement, we made some great upgrades to our IDP platform over the summer. As you know, Infrrd empowers our clients with solutions for intelligent work that take advantage of end-to-end process automation, a critical part of which is automated data extraction. Our intelligent document processing (IDP) platform leverages a machine learning (ML) model that any user can use to create a powerful data extraction solution delivering efficiencies and cost savings far beyond the reach of standard optical character recognition (OCR) systems. Data extraction can get extremely messy and necessitates human-in-the-loop intervention to improve the efficiency of the algorithm, although ensuring 100% accuracy can be a challenge. Our improved Audit feature addresses this aspect by making it easier for users to extract and review corrected data.
Now, when our clients sign up to use our IDP platform, they can access the ML model through what we call the Project Module, and train the model to extract data per their custom needs. With this new project module, our clients can manage classification, correction configuration, data extraction and SLA — all together in one place. This gives our business users the flexibility to change key configurations and change their data extraction process without involving IT.
The IDP platform was also designed to process structured and semi-structured documents separately, with users specifying the document types. For instance, users would need to process invoices separately from tax forms. Now, through the group model, the platform can automatically separate documents of different, predefined types without requiring further user input to carry out the data extraction.
Simplifying error correction
Within the same Project module, users can create correction queues to fix any errors in the extraction.
Automatically splitting long documents containing many individual documents scanned together into individual documents has already been a feature of our product. This has helped clients save significant time by not having to split documents manually and let the mailroom users scan all the documents in one bulk scan.
With this release, we have advanced this feature by enabling clients to visually correct any errors in page splitting. This speeds data extraction in documents that need to be reviewed for page splitting before data can be extracted from them. Additionally, our ML algorithms learn from each of these corrections to improve future split accuracy.
Similarly, by configuring the Copy Over feature when correcting the data extracted from a group of related documents, the user can replicate values for common fields present across different document types. Apart from saving the user’s time, this feature also eliminates the need to re-extract data or manually populate fields with data, which could introduce further errors.
Where the correction involves reclassifying a document within the group model, the Subflow feature can be used to reprocess the document. Utilizing these features can allow clients to maximize their return on investment, in terms of resources employed for correcting the extracted data. In its earlier version, the Infrrd platform displayed reports within the system, which are now available to users via APIs, enabling data visualization in any preferred format. Users tasked with making corrections can leverage and integrate this feature into their dashboards and track data extraction information.
Guaranteeing 100% accuracy via auditing
The other key component of the Infrrd IDP platform is the Audit Module, designed to allow sampling extracted (and corrected) data and verify that the data extracted is 100% accurate. We understand that no matter how trained or experienced the correcting team might be, they may need to make judgment calls about which other users may not always know. Through the Audit module, an administrator can check the correcting user’s documents and review their results. Data auditors can also download a report that can be leveraged to improve the ML model as well as the correction methodology.
The audit feature is a must-have tool for supervisors with large data entry teams. Clients can now configure auditing parameters for supervisors to randomly check documents processed by some users. It is a great way to introduce new data processors into the system and watch over their shoulders as they come up to speed and make teething errors.
Our IDP application can be managed easily by customers through a customized control center that allows them to add and replace documents, review and redefine data extraction, and provide inputs for improving the platform’s performance, translating into a lower risk of inaccurate analysis. Such a data extraction technique is a lasting solution that is also adaptable to newer document formats and data fields. Ultimately, it can offer enhanced data management value for business users who no longer need data scientists to handle the platform for them. If you want to understand how Infrrd can impact your business, take a look at our platform.
Infrrd’s spirit of innovation
At the end of the day, all of us at Infrrd are committed to empowering your teams with smarter and more efficient solutions which get the job done with the highest level of accuracy. Our dedicated engineers and developers are constantly working to improve our platform further, adding features that grow in tandem with your organization and its changing needs. We are here to solve not only today’s business challenges, but also forestall the challenges of tomorrow.
Originally published at https://www.infrrd.ai.
