Data Extraction (OCR/ICR)

Automated data entry is a core element of PHX Terminal’s value proposition: it removes manual input, reduces errors, and saves time by populating forms and transferring data between disparate applications. This page covers the technologies that extract data and the mechanisms that move it.

Intelligent data extraction

Several complementary technologies underpin extraction:

Technology	Purpose
OCR (Optical Character Recognition)	Converts images, scanned documents, and PDFs into editable, searchable text — essential for digitizing physical legal documents.
NLP (Natural Language Processing)	Extracts meaning, sentiment, and key information from unstructured text such as correspondence and case notes. See LLMs & NLP.
Machine Learning / AI	Adapts to varying data formats and continuously improves extraction accuracy through iterative training.
ICR (Intelligent Character Recognition)	An advanced form of OCR that extracts data from physical documents including handwritten text, recognizing diverse handwriting styles and fonts.

Tip

Organizations using AI-driven data-cleansing tools have reported a 40% improvement in data accuracy — a direct driver of higher data integrity and fewer costly mistakes in legal documents, billing, and records.

Cross-application data transfer

Once data is extracted, PHX Terminal moves it across the firm’s software estate using several mechanisms, chosen by what each target application supports:

• Direct API integrations — the primary method for structured data exchange where modern APIs exist.
• Automation platforms (e.g., Zapier or Make) — to create data flows between applications without extensive custom code.
• Smart spreadsheet functions — for automated manipulation and transfer within spreadsheet environments.

Programmatic UI-level injection

For the many legal applications that lack robust APIs — especially legacy and virtualized software — PHX Terminal injects data directly at the UI layer. This is achieved through:

• Accessibility APIs for native application interaction (e.g., Microsoft UI Automation, Apple Accessibility API).
• Specialized UI automation frameworks such as UI Automator (Android) and WinAppDriver / FlaUI (Windows).

These frameworks let the bots interact with UI elements at a programmatic level, enabling precise data injection across diverse desktop applications. Combined with AI Computer Vision, this gives PHX Terminal a reliable path to both read from and write to virtually any application — the technical basis for the hover opaque application acting as a universal adapter.

flowchart LR
  SRC["Documents & screens<br/>scans · PDFs · handwriting · text"]
  subgraph EX["Extraction layer"]
    OCR["OCR"]
    ICR["ICR — handwriting"]
    NLP["NLP"]
    ML["Machine learning / AI"]
  end
  SRC --> EX
  EX --> DATA["Structured data"]
  DATA --> TRANSFER{"What does the<br/>target support?"}
  TRANSFER -->|"modern API"| API["Direct API integrations"]
  TRANSFER -->|"no-code flows"| ZAP["Zapier / Make"]
  TRANSFER -->|"spreadsheets"| SS["Smart spreadsheet functions"]
  TRANSFER -->|"legacy / no API"| INJ["UI-level injection<br/>Accessibility APIs · UI Automator · WinAppDriver / FlaUI"]
  API --> DEST["Destination applications"]
  ZAP --> DEST
  SS --> DEST
  INJ --> DEST

Extraction turns documents and screens into structured data; the platform then routes it to each target by whatever that application supports — from modern APIs down to UI-level injection for legacy software.

Where it fits

Extraction feeds the rest of the automation pipeline: Computer Vision recognizes the document or screen, OCR/ICR/NLP extract and structure the content, and Intelligent Process Automation carries the structured data into the destination systems — all under human oversight.