Glossary

Shared vocabulary used across workflow, tool, and delivery pages so you can move between sections without reconciling conflicting terms.

Pipeline modules and data layers

Term	Meaning
Parser	Extraction stage (implementation repo: gen-sci-data) that turns papers into reusable text blocks, figures, captions, and metadata under the Raw Database.
Processor	Quality stage that cleans Parser output into the Processed Database so downstream steps see stable text and fields.
Designer	Delivery stage that organises stable input into full-text, structured, and multimodal outputs under the Designed Database.
Raw Database	Layer retaining originals and Parser bundles for traceability (default root often data/Raw Database/; legacy docs may show a raw/ segment).
Processed Database	Layer holding cleaned intermediates and processing reports (default root often data/Processed Database/).
Designed Database	Designer output root: per-doc_id folders with *.content.json, *.structure.json, *.assets.index.json, and images/. Override the path with --output_dir.

Output and field concepts

Term	Meaning
full_text	Human- and machine-readable full-document delivery (JSON + parallel plain text as documented).
datamining	Section-oriented structured JSON aimed at extraction, retrieval, and storage.
multimodal	Indexes and assets linking in-text figure references to image files and captions.
doc_id	Stable per-document identifier used across all layers.
Caption	Figure or table descriptive text, often critical for review and mining.
Asset anchor	Stable link between a mention in text and a concrete file in the delivery tree.
processing_report	Batch-level summary with status and quality signals for review.
Academic atomicity	Design rule: preserve scientific meaning, structure, and context during cleaning unless an explicit, reviewed rule says otherwise.

See also: Home · Quick Start