PDF Text Recognizer Pro

A desktop application for PDF text extraction, font-aware layout analysis, and citation detection, designed for academic and technical documents.

---

Features

Core Capabilities

• Text extraction with exact span-to-PDF coordinate mapping
• Font-aware layout analysis (font name, size, style)
• Superscript & bracket citation detection engine
• Bibliography parsing with false-positive suppression
• Modern GUI with synchronized PDF/Text views

---

Architecture

PDF (pdfplumber)
   ↓
PageData / LineData / CharData
   ↓
Citation Channels (Superscript / Bracket)
   ↓
Fusion Engine (confidence scoring + filtering)
   ↓
RefEntry / Occurrence
   ↓
GUI (PDF ↔ Text ↔ Citation sync)

• All GUI updates run on the main thread
• Background tasks are cancelable via job-id invalidation
• Image rendering uses LRU cache to cap memory usage

---

Citation Engine Design

Detection Channels

• Superscript channel: geometric + font-size based detection
• Bracket channel: [n], (n) style inline citations

Bibliography Handling

• Strict line-head ID matching (^\s*(\[(\d+)\]|(\d+)\.))
• Year-number filtering (1900–2099) to prevent ID pollution
• max_id_multiplier false-citation upper bound

Soft Constraint System

• When bibliography is reliable (≥ N entries), unlinked citations are penalized but not discarded
• Small or missing bibliographies automatically disable penalties

---

Reliability & Safety

• Thread-safe background execution (no Tk access in workers)
• Job ID mechanism prevents stale callbacks from overwriting state
• PDF handle caching with guaranteed release on exit/errors
• Debug reports never include raw document text

---

Build & Run

Run from source

python app_gui.py

Windows executable

• Prebuilt executable: dist_exe/PDFTextRecognizer.exe
• No Python environment required

---

Testing & Verification

Included test scripts:

• test_citation_improvements.py – citation logic validation
• reconciliation_check.py – configuration & import-path verification

All tests must pass before deployment.

---

Scope & Non-Goals

• Not intended for OCR (scanned PDFs)
• Not a reference manager or citation formatter
• Focused on structural correctness and traceability, not heuristics-only extraction

---

License

MIT LICENSE