Configurable requirements for tasks in document transcription by neural networks
Configurable neural networks for document transcription address the issue of insufficient granularity by allowing precise annotation selection, improving efficiency and reducing resource usage.
DE102025147261A1Pending Publication Date: 2026-05-21NVIDIA CORP
View PDF 0 Cites 0 Cited by
Patent Information
- Authority / Receiving Office
- DE · DE
- Patent Type
- Applications
- Current Assignee / Owner
- NVIDIA CORP
- Filing Date
- 2025-11-14
- Publication Date
- 2026-05-21
AI Technical Summary
Technical Problem
Existing neural networks for document transcription lack configurability, leading to insufficient granularity in defining content, resulting in unnecessary processing costs and resource utilization due to inclusion or exclusion of unwanted annotations.
Method used
Neural networks are trained to generate document transcriptions with specific annotations as specified by configurable prompts, allowing users to select desired annotations and exclude unwanted ones, using mixed training data and architectures like transformer-based vision-encoder-decoders.
Benefits of technology
This approach enhances the accuracy and efficiency of document transcription by optimizing resource use and reducing processing overhead, ensuring precise annotations for downstream tasks.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure 00000000_0000_ABST
Abstract
Devices, systems, and techniques for generating a document transcription of a document image. In at least one embodiment, one or more neural networks generate a document transcription of a document image according to a configurable combination of annotation types input into the one or more neural networks. The document transcription can contain respective annotations of the annotation types for corresponding parts of the content included in the document transcription.
Need to check novelty before this filing date? Find Prior Art