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)
87 lines
1.9 KiB
YAML
87 lines
1.9 KiB
YAML
# Scientific Paper extraction profile
|
|
# Matches academic papers from arXiv, journals, conference proceedings
|
|
name: scientific_paper
|
|
description: Academic papers from arXiv, journals, conference proceedings
|
|
priority: 30
|
|
|
|
match:
|
|
all:
|
|
- any:
|
|
- text_contains:
|
|
patterns: ["Abstract", "References", "doi:", "arXiv:", "Bibliography"]
|
|
- heading_matches:
|
|
pattern: "^(Abstract|Introduction|References|Bibliography)"
|
|
- any:
|
|
- structural:
|
|
has_table: false
|
|
has_form_field: false
|
|
has_math: true
|
|
page_count: null
|
|
- structural:
|
|
has_table: false
|
|
has_form_field: false
|
|
has_math: false
|
|
page_count:
|
|
min: 4
|
|
max: 50
|
|
none:
|
|
- text_contains:
|
|
patterns: ["Invoice", "Receipt", "WHEREAS", "NOW THEREFORE"]
|
|
|
|
extraction:
|
|
reading_order: xy_cut
|
|
table_detection: default
|
|
readability_threshold: 0.5
|
|
include_invisible: false
|
|
include_headers_footers: false
|
|
force_ocr: false
|
|
min_block_chars: 0
|
|
|
|
fields:
|
|
title:
|
|
type: string
|
|
extraction:
|
|
region: top_quarter
|
|
pick: largest_font
|
|
parse: string
|
|
|
|
authors:
|
|
type: array
|
|
extraction:
|
|
region: top_quarter
|
|
pick: nearest_below
|
|
after_heading: title
|
|
fallback: []
|
|
|
|
abstract:
|
|
type: string
|
|
extraction:
|
|
near: ["Abstract"]
|
|
region: top_half
|
|
parse: string
|
|
|
|
doi:
|
|
type: string
|
|
extraction:
|
|
regex: "doi[:\\.]\\s*(10\\.\\d{4,9}/[\\w\\-\\._;()/:]+)"
|
|
parse: string
|
|
|
|
journal:
|
|
type: string
|
|
extraction:
|
|
region: top_eighth
|
|
pick: first
|
|
parse: string
|
|
|
|
publication_date:
|
|
type: date
|
|
extraction:
|
|
near: ["Published", "Received", "Accepted"]
|
|
parse: date
|
|
|
|
references:
|
|
type: array
|
|
extraction:
|
|
region: bottom_half
|
|
after_heading: References
|
|
fallback: []
|