80dbf0f703
- Add profile source modules: apply_profile, extraction, extraction_loader, field_extractor, match_eval - Add profiles CLI subcommand (profiles_cmd.rs) - Update all 9 built-in profile YAMLs (invoice, receipt, contract, scientific_paper, slide_deck, form, bank_statement, legal_filing, book_chapter) - Add 50 invoice fixture PDFs - Add 2 receipt fixture PDFs Part of: pdftract-3a310 (Phase 7.10 coordinator)
59 lines
1.2 KiB
YAML
59 lines
1.2 KiB
YAML
# Slide Deck extraction profile
|
|
# Matches PowerPoint / Keynote / Google Slides exports as PDF
|
|
name: slide_deck
|
|
description: PowerPoint / Keynote / Google Slides exports as PDF
|
|
priority: 15
|
|
|
|
match:
|
|
all:
|
|
- structural:
|
|
has_table: false
|
|
has_form_field: false
|
|
has_math: false
|
|
page_count:
|
|
min: 3
|
|
max: 200
|
|
- any:
|
|
- text_matches:
|
|
pattern: "^Slide \\d+$"
|
|
- text_contains:
|
|
patterns: ["slides", "presentation"]
|
|
none:
|
|
- text_contains:
|
|
patterns: ["Abstract", "References", "WHEREAS", "Invoice"]
|
|
|
|
extraction:
|
|
reading_order: xy_cut
|
|
table_detection: default
|
|
readability_threshold: 0.6
|
|
include_invisible: false
|
|
include_headers_footers: false
|
|
force_ocr: false
|
|
min_block_chars: 5
|
|
|
|
fields:
|
|
title:
|
|
type: string
|
|
extraction:
|
|
region: top_half
|
|
pick: largest_font
|
|
parse: string
|
|
|
|
presenter:
|
|
type: string
|
|
extraction:
|
|
region: top_half
|
|
pick: largest_font
|
|
parse: string
|
|
|
|
date:
|
|
type: date
|
|
extraction:
|
|
near: ["Date"]
|
|
parse: date
|
|
|
|
slide_titles:
|
|
type: array
|
|
extraction:
|
|
pick: largest_font
|
|
fallback: []
|