Vendor Analysis
published on Apr 09, 2020
Report Overview:
NelsonHall’s ABBYY FlexiCapture document cognition benchmark report is a functional “lab” test evaluation of the ABBYY FlexiCapture document cognition platform.
Who is this Report for:
NelsonHall’s ABBYY FlexiCapture document cognition benchmark report is a functional “lab” test evaluation of the ABBYY FlexiCapture document cognition platform, designed for:
- Sourcing managers monitoring the capabilities of existing suppliers of Document Cognition platforms and identifying vendor suitability for document cognition RFPs
- RPA and intelligent automation center of excellence personnel evaluating document cognition platform capability
- Vendor marketing, sales, and business managers looking to benchmark their platforms against their peers
- Financial analysts and investors covering intelligent automation and Document Cognition platforms.
Scope of this Report:
The report provides a comprehensive and objective analysis of ABBYY FlexiCapture’s capabilities, including:
- Designing the document cognition models
- Document Ingestion
- Document verification
- Testing.
This report provides the quantitative results of the SmartLabTest comparing the platform’s performance for each document type. This report includes:
- The SmartLabTest results
- Analysis of the platform’s strengths & weaknesses
- Identification of the key features of each platform
- An evaluation of comparative platform maturity
Key Findings & Highlights:
This NelsonHall vendor assessment tested ABBYY FlexiCapture’s capabilities in ingesting and interpreting:
- Structured documents such as mortgage applications and ACORD filings
- Semi-structured documents such as invoices and purchase orders
- Highly unstructured documents such as resumes.
The KPIs assessed for each document type included:
- Proportion of fields correctly recognized
- Accuracy of extraction of recognized fields
- Proportion of fields overall that 100% accurate and require no manual intervention.
FlexiCapture uses a mixture of algorithms to classify and capture document types and content, including images accurately. Algorithms used to understand documents include:
- Convolution Neural Networks (CNN)
- Recurrent Neural Networks (RNN)
- Gradient boosting on deep decision trees
- Client-side supervised learnings
- Unsupervised learnings.
ABBYY was most successful when used with the highly structured documents in the test. As expected, the platform handled changes in the image quality reasonably well, with no significant loss in the percentage of fields recognized or the accuracy of said fields with noisy, skewed, and rotated images, with the accuracy of recognized fields falling when assessing documents at 100 DPI.