(888) 575-9299
Login

RESOURCES

Whitepapers

IDR Systems

Benefits of Intelligent Document Recognition

IDR Systems (Intelligent Document Recognition), in areas like invoice capture, typically offer a return on a user’s investment in less than 12 months—mainly as a result of reduced labor costs related to data entry. In a recent survey on invoice processing conducted by IAPP-TAWPI, 75% of respondents said that automated data capture technology reduced their invoice processing costs by up to 25%, and another 15% estimated it reduced costs by 25-75%. In addition, respondents cited increased visibility of their processes, faster invoice turnaround time, and fewer errors as major benefits of their IDR implementations.

Applications of IDR Systems

Of course, while invoice capture is certainly the most prevalent application for IDR technology today, it is certainly not the only one. Other areas being addressed include patient records and explanation of benefit (EOB) forms in the healthcare industry, transcripts in higher education, and reports in the financial services industry. The use of traditional [less intelligent] forms processing software based primarily on OCR/ICR technology also continues to be popular for documents like insurance claims, order forms, and tax returns.

Benefits of Automated Data Capture

While automated data capture is certainly technology with proven benefits, its application can be doomed by shortfalls in two areas: low accuracy rates and extensive set-up requirements. Low accuracy can create data correction and validation requirements, as well as downstream data errors that end up minimizing or nullifying the cost benefits of the system. Extensive set-up can put a burden on systems administrators and/or involve expensive consulting bills that also eat into an application’s ROI.

OCR Accuracy Comparison

Black & White Color
SoftWorks AI 11,650 4,905
Leading OCR Engine 10,151 4,421
Accuracy Ratio SoftWorks AI is 15% better SoftWorks AI is 11% better

 

The holy grail of automated data capture is software that can, out-of-the-box, produce results accurate enough to reduce data entry costs by at least 50%. In an IDR application, like invoice capture, the industry standard for a successful application is typically around 80% accuracy on a field level. This means that four out of five data fields are captured accurately and require no human intervention, except maybe a quick approval. In traditional forms processing (non-IDR) applications, field-level accuracy rates are often well over 90%.

In a world where the use of document imaging and automated data capture technology continues to grow (as the technology matures and users seek to increase the efficiency of their paper-based transactions to keep up with the speed of their electronic transactions), ideally a document capture software package should be able to address both IDR and traditional forms processing requirements. It should also be easy to get up and running on new document sets, as well as additions to existing document sets, to keep administrative costs down. And, finally, it should produce results that meet or exceed industry standard accuracy rates.

Trapeze Software

SoftWorks AI’s Trapeze represents a software package that meets these requirements. Out-of- the-box, Trapeze offers 90% field-level accuracy in IDR applications. It can also be used for traditional, template-based forms processing. And in both cases, the software can utilize information entered by data entry operators to increase its accuracy over time.

SoftWorks AI brings a unique background to automated data capture. Its legacy as a document compression specialist has enabled it to develop some strong image processing techniques, which enable Trapeze to produce cleaner text than most applications, which increases the accuracy of OCR/ICR. SoftWorks AIhas also developed Expert Systems, or sets of rules, around specific capture applications that enable its customers to get their systems up in and running in a minimal amount of time.

 

SoftWorks AI’s Trapeze provides automated file compression, data extraction and OCR solutions.

Trapeze’s Accuracy

Tests have shown that Trapeze is 10-15% more accurate, on a word-level, than a leading OCR/ICR application. This means that SoftWorks AI customers can potentially reduce their data entry requirements an additional 10-15% over what they could with competitive software. For an organization with 10 to 15 key-entry personnel, this equates to the reduction of one full-time employee. Increased accuracy also has other benefits, like reducing downstream data mistakes, which can often be expensive to fix, especially in markets dealing with financial transactions.

SoftWorks AI’s high accuracy is due in part to its ability to leverage the mixed raster content (MRC) methodology it developed for image compression. MRC involves separating the textual elements of document images from the graphics and the background. This enables each segment of a document to be compressed with different technology to create optimal file sizes while maintaining image quality.

Being able to successfully separate text can be utilized to produce a cleaner set of data to feed an OCR/ ICR engine. This is especially helpful for low-contrast images, on which the text is not readily discernible from the background, and noisy images, which might include mark-ups and background information like watermarks.

Trapeze’s Color Advantage

SoftWorks AI’s experience compressing color images also gives Trapeze unique advantages over other image processing technologies, which primarily work with bi-tonal or gray scale images. Being able to leverage the greater amount of data included in color enables Trapeze to make more informed decisions, which can prove especially valuable when dealing with certain types of low contrast images— like those where the text and background may be different shades of the same color.

And with color document scanners now priced equally with black-and-white models, there is little reason for users not to take advantage of color.

Trapeze’s Image Advantage

SoftWorks AI also has strong technology in the areas of deskewing, image de-warping, and pattern recognition that can be valuable in data capture applications. Deskewing is the ability to straighten out images and data so they can be recognized correctly. De-warping involves normalizing areas of an image that may be stretched or otherwise deformed due to the way the image was captured. SoftWorks AI’s pattern recognition expertise was honed through its work with JBIG2 compression. In an OCR/ICR application, it can be leveraged to increase the recognition rates of handwritten characters.

SoftWorks AI’s Trapeze represents a powerful and accurate automated data capture platform that can be utilized to significantly reduce data capture and document processing costs. Out-of-the-box, with little customization, Trapeze can produce higher than industry-standard accuracy rates on a variety of document types. This includes structured forms like order forms, surveys, and tax returns, as well as semi-structured forms like invoices, transcripts, patient records, and financial transaction documents.

SoftWorks AI’s template-less approach to semi-structured forms and its Expert Systems modules, which include application-specific intelligence, enable Trapeze to be set up in a matter of days or weeks, vs. competitive applications, which often require months of custom- ization. And SoftWorks AI’s advanced image processing techniques enable Trapeze to produce results that are 10-15% more accurate that than those of leading OCR engines. This combination of high accuracy and ease-of-setup makes Trapeze a compelling application for most automated document and data capture requirements and enables it, in most cases, to provide a return-on-investment, in well less than a year.

Trapeze Software: Ease of Use

Of course, producing more accurate character recognition alone does not make an effective data capture application. The software has to know what to do with the data it receives. To facilitate this process, SoftWorks AIhas introduced a series of Expert Systems, or rules related to processing data for specific applications.

Trapeze’s Expert Systems are associated primarily with IDR applications. That’s because for traditional forms processing applications, the data capture process is a lot more straightforward. Traditional forms processing is typically applied to structured forms documents where the same type of data appears consistently in the same place.

Structured forms can typically be addressed using templates. Templates are set up by marking/highlighting and labeling data fields on an image. The template tells the forms processing application to look in the same spot for the same type of data each time a form with the same design appears.

IDR is typically applied to semi-structured forms. These are documents that contain the same type of data, but which can appear in different places and even be labeled differently. Semi-structured forms also often include tables of information, which can vary in length and construct.

Applying templates to semi-structured forms can be a complex and time-consuming process. In an invoice processing application, for example, it could potentially mean setting up templates for each different vendor an A/P department deals with and even smaller organizations often deal with hundreds of vendors that are constantly changing. For a college or university, it could involve setting up templates for each institution transfer students come from. A hospital could require a separate template for each insurance company, and so on. Some proposed data capture applications have never made it to the production stage because users have been so overwhelmed by the task of setting up the templates.

Trapeze’s Template-less Approach

Trapeze takes a template-less approach to semi-structured forms. This involves applying OCR/ICR to the entire document and then feeding those results to one of its Expert Systems. The Expert Systems look for specific keywords and data constructs and apply rules to automatically classify data and documents. Because there is no field location information to assist in the recognition process, the accuracy of OCR/ICR data takes on additional importance in template-less applications, which creates an advantage for Trapeze over most applications.

Utilizing its Expert Systems modules, without any customization, Trapeze can achieve field level accuracy rates of 90% on semi-structured forms, or 10% higher than the accepted accuracy rates of most applications in production. Trapeze also has the ability to capture line-item detail, such as item numbers, quantities, and price per item on invoices, or transaction details on financial reports, at an accuracy rate of close to 80% per line item. Because of the complex construction of line items, such as their running on for multiple lines, or even on to multiple pages if a description is included, many capture applications do not even address them. For increased accuracy, Trapeze offers functionality like checking the total amount of an invoice by adding the totals of the line items.

Being able to read line items can result in an increased rate of straight-through processing documents which do not need to be seen by a human operator. Reading line items enables for more complete auto-matching against supporting documentation, such as purchase orders, which ensures accuracy of data being extracted and posted to a back end system. Three-way matching can help further reduce document processing costs.

Applications of Trapeze

Although Trapeze can be integrated with back-end systems to enable functionality such as three-way matching, it does not have to do this to achieve its high-accuracy rates. Many other capture systems require some sort of database matching to achieve acceptable recognition rates. Integrations with third-party systems to access data can be costly and time consuming, as well as open up back-end systems in a way that could compromise security.

Setting up a Trapeze IDR application typically involves running several hundred sample documents through the software, mainly to adjust the image processing components, so that the cleanest possible characters are fed to the OCR engine. Sample documents are also all that is needed to invoke Trapeze’s auto-document classification capabilities. An administrator can group specific types of document images together from the sample set, and Trapeze has the ability to learn the characteristics of each group and compare them to the characteristics of images being captured in the future.

Trapeze’s Smart Auto-Classification

In a mortgage file application, for example, a user might scan 100 files as a sample set, setting up separate groups for deeds, W-2s, credit reports, closing forms, etc. Trapeze should then know enough to automatically classify the contents of any mortgage file captured thereafter. Patient records are another application for auto-classification, especially as healthcare organizations transition to electronic medical records systems to meet government requirements.

Any unrecognized document is placed in a queue for manual classification. Trapeze learns from this process, and the next time it encounters an image with the same characteristics, it should be able to automatically classify it. These same learning capabilities are applicable in data extraction applications, for both structured and unstructured forms. Trapeze has a browser-based validation and data entry interface that enables operators to select an area on an image and use the OCR/ICR results from this selected area to automatically populate a data field. The next time Trapeze encounters a form with a similarly structured data field in the same location, it should automatically capture the data.

Bottom Line: High Accuracy, Low Cost

The bottom line is that SoftWorks AI’s Trapeze represents a powerful and highly accurate automated data capture platform that can be utilized to significantly reduce data capture and document processing costs. Out-of-the-box, with little customization, Trapeze can produce higher than industry-standard accuracy rates on a variety of document types. This includes structured forms like order forms, surveys, and tax returns, as well as semi-structured forms like invoices, transcripts, patient records, and financial transaction documents.

SoftWorks AI’s template-less approach to semi-structured forms and its Expert Systems modules, which include application-specific intelligence, enable Trapeze to be set up in a matter of days or weeks, vs. competitive applications, which often require months of customization. And SoftWorks AI’s advanced image processing techniques enable Trapeze to produce results that are 10-15% more accurate that than those of leading OCR engines. This combination of high accuracy and ease of setup makes Trapeze a compelling application for most automated document and data capture requirements. In most cases, it can enable achievement of ROI in less than a year.

Request Demo Download Whitepapaer

CLIENT SUCCESSES

See how SoftWorks AI can help your organization

Contact Us