Kensho Extract

Kensho Extract is a machine learning-powered document processing tool that transforms unstructured PDFs and text files into structured, machine-readable formats. It enables fast, accurate extraction of text, tables, and key-value pairs, streamlining workflows for data analysis, enrichment, and integration across enterprise systems.

Features

• Optical Character Recognition (OCR) for scanned documents
• Hierarchical document structure recognition
• Advanced table extraction with support for merged cells and headers
• Figure and chart data extraction (e.g., bar charts)
• REST API for high-throughput document processing
• JSON output with detailed segmentation (titles, paragraphs, tables, footnotes)
• Support for multiple output models (hierarchical, general)
• UI for manual review and correction
• Integration with Kensho NERD and Link services
• Toolkit for markdown conversion, section organization, and visual formatting

Capabilities

• Convert complex PDFs into structured data for analysis
• Extract and standardize financial metrics, periods, and currencies
• Identify and isolate tables, figures, and key-value pairs
• Maintain page layout for multilingual translation workflows
• Enable downstream NLP and machine learning applications
• Automate data onboarding and reduce manual copy-paste errors
• Handle inconsistent formatting and messy layouts
• Support batch and real-time processing for large-scale operations

Benefits

• Saves time and resources in document processing
• Improves data quality and accessibility for analytics
• Reduces reliance on manual data entry and spreadsheet manipulation
• Enhances integration with internal databases and external knowledge bases
• Enables faster decision-making with structured insights
• Scales across teams and departments with minimal setup
• Supports compliance and audit workflows through accurate data extraction
• Boosts productivity in financial, legal, and research environments