In October, ABBYY, an Ashling / Reveal Group partner, hosted its annual ABBYYY ASCEND Webinar. ABBYY unveiled a range of groundbreaking advancements, signaling the next wave of innovation for the coming year. Themed “Innovating AI, Simplifying Success,” the event underscored ABBYY’s commitment to making AI-driven Intelligent Document Processing (IDP) solutions more accessible and practical. Below, we summarize some of the most impactful new features designed to elevate user experience and simplify the deployment of ABBYY’s transformative AI capabilities.
Phoenix 1.0
This year, Phoenix 1.0 was introduced at the ABBYY ASCEND Webinar. A multimodal AI model for document processing. Unlike broad LLMs, ABBYY Phoenix was purpose-built to meet the nuanced text and visual requirements of document-based processing, focus on speed and accuracy, and minimize resource consumption to reduce operational costs. ABBYY has donated its AI research to the broader community for the past four years through NeoML, its open-source machine learning framework for image preprocessing, classification, document layout analysis, OCR, and data extraction from structured and unstructured documents.
Auto-Labeling
Phoenix is the engine behind Vantage’s new auto-labeling, a powerful feature to expedite labeling to onboard new document types.
Leveraging the capabilities of the Phoenix LLM, ABBYY Vantage can now automatically label documents, reducing model development time by half. This is achieved through the creation of one-time document field presets that adapt to various model requirements. Once initial pre-labeling is complete, Vantage and Phoenix apply these presets across similar documents within the model, significantly streamlining the process.
A similar capability was introduced to UiPath Communications Mining this year, and it has helped Ashling Partners / Reveal Group significantly accelerate the automation of contact center emails. We have found that human verification remains essential, but auto-labeling offers substantial efficiency gains.
Prompt-Based Activity
Establishing connections and prompts between third-party LLMs, such as GPT, Pi, and Gemini with Vantage, which previously required custom JavaScript code stages. ABBYY’s low-code interface now includes a new prompt-based activity, allowing low-code developers to configure and execute prompts directly within the standard user interface. Results are returned in real time, removing the need to exit the activity screen or deploy the solution to test.
Ashling Partners / Reveal Group and ABBYY have been working closely on this capability, which has allowed us to solve otherwise impossible unstructured document processing challenges. A simplified example, which centers around extracting 300+ fields from lease documents, was shown on the ASCEND Webinar.
Quality Analytics Dashboard
While ABBYY Vantage has many improvements over the ABBYY Flexicapture, easily accessible business dashboarding has lagged. The Ashling Partners / Reveal Group team typically wraps ABBYY in orchestration and automation platforms, such as RPA, to connect to Vantage’s data warehouse via API. We then visualize highly customized reports using external tools such as PowerBI, Tableau, etc.
Now, ABBYY is introducing a quality analytics dashboard that will give users easily accessible real-time insights. Some key reports include:
- Straight-Through Processing: Tracks documents processed without human intervention.
- Document Classifier %: Displays the classification success rate, helping users identify which classification types are underperforming and may need more training.
- Field Extraction Accuracy %: Shows the accuracy rate of field extractions, highlighting fields that may require additional training for better extraction performance.
- Data Drilldowns: Allows users to explore niche data points such as Image Type, Image Source, and Image DPI.
Each contains further breakdowns, such as document type, image details, and documents that don’t need correction.
Timeline
Finally, Reveal Group alumni John Knisley presented new features of ABBYY Timeline. These enhancements focus on the push to enable more data-driven decision-making and enhanced scalability and reliability of the platform. Now infused with AI-powered analysis and predictive modeling, the enhanced Timeline platform has scaled up from handling millions to billions of event-driven data points. This scalability allows Timeline to harness the power of process intelligence to monitor, predict, simulate, and alert for complex event-driven data in real time.
For full details, watch the ABBYY 2024 ASCEND Webinar
Continued Learning
Are you interested to learn how ABBYY can enhance your document automation capabilities?
Author: Damien Binda
Business Analyst Manager at Ashling Partners & ABBYY Vantage Certified Developer
