Working directory for the reduced manuscript ("text-mining-only" scope, see
outline.txt). Contents copied out of ../HAMLET on 2026-08-18, keeping only
what the outline's scope decision keeps IN.
framework/-- the agentic extraction pipeline:agents/(biological, technical, experimental-design, integration, normalization agents),core/(extractor, prompts, field mappings, LLM wrapper),normalization/(ontology graph, SapBERT/FAISS index, synonym handling),validation/(structural rules, gleaning),tests/,docs/, plusmain.py,config.yaml,prompts_reference/. This is Section 2 / M&M 5-8 of the outline. Source:HAMLET/src/agentic-metadata/agentic-metadata/.benchmark/-- SDRF-derived ground truth and matching code (I9/I10/I11, Section 3-4, M&M 10):benchmark_data/(matched, sdrf_golden, test_set, new_test_set,run_sdrf_benchmark.py,sdrf_to_golden.py,annotation_to_golden.py, plotting scripts),run_bench.sh.corpus/-- human annotation corpus and PRIDE-comparison materials (I2-I6, I13, Section 1 / Figure 1 / Figure 6, M&M 1-3):Figure1_source/--data/Select_27_Pubs,data/GoldenAnnotations(Annotations,HarmonizedHuman,SingleHuman,MultiHuman,GPT,Representative20,KaggleHoldout),field_crosswalk_table.csv,make_figure1.py/make_figure1_v2.py, prioroutput/output_v2.pride_metadata_comparison/--HamletPXDs.csv(I13, n=299) and the fullpxd_lists/directory.pride_census/--pride_survey.pyonly. The raw crawl (pride_survey/, ~6.1 GB) was not copied; seeNOTE.mdthere.
ontology_refs/-- local CV/ontology source files referenced byframework/normalization/:unimod/,taxonomy/,taxid_lists/,sdrf-terms.csv(I14). CL, CLO, UBERON, Mondo, PSI-MS CV, and DOID are fetched byframework/normalization/download.py, not stored here.docs/--Metadata_paper_v2.1-2.pdf(current full draft, superseded by the reduced outline) for reference.outline.txt-- the reduced manuscript outline (moved, not copied, from the agentic-metadata submodule).