paper-digest-single

name	paper-digest-single
description	Digest a single scientific paper into a structured summary. Accepts a PubMed ID, DOI, arXiv ID, or local PDF path. Fetches abstract and metadata from PubMed/CrossRef APIs, extracts key findings, methods, results, and clinical implications, and saves a structured markdown report.

Paper Digest Single

Purpose

Extract and synthesize key information from a single scientific paper (PubMed, arXiv, CrossRef, or local PDF) into a structured, human-readable markdown report. Ideal for rapid literature review, evidence synthesis, and knowledge base building.

When to Use / When NOT to Use

Use when:

• You need to quickly understand a single paper's main contributions
• Building a personal knowledge base from specific papers
• Preparing literature review sections that cite specific findings
• Evaluating clinical or research evidence for a specific topic
• You have a PubMed ID, DOI, arXiv ID, or PDF file ready to process

Do NOT use when:

• You need to compare multiple papers at once (use pubmed-search or preprint-tracker instead)
• You want to search for papers on a topic (use pubmed-search first)
• You need full-text PDF analysis beyond abstract-level summaries
• Processing confidential or restricted-access papers

Expected Inputs & Outputs

Inputs (choose ONE):

• --pmid INT — PubMed ID (e.g., 35769483)
• --doi STRING — Digital Object Identifier (e.g., 10.1038/s41586-022-04826-7)
• --arxiv STRING — arXiv ID (e.g., 2301.12345)
• --pdf PATH — Local PDF file path

Outputs (in --outdir):

• paper_digest_<identifier>.md — Structured markdown report
• paper_metadata.json — Complete metadata (authors, MeSH, keywords, journal, etc.)
• extraction_log.txt — Processing notes and warnings

Procedure

1. Fetch Metadata

   • PubMed (via eFetch XML API) or CrossRef API for DOI
   • Extract: title, authors, journal, year, abstract, MeSH terms, publication type

2. Structured Text Extraction (rule-based NLP)

   • Background: opening sentences (study motivation, context)
   • Methods: sentences with method keywords (RNA-seq, CRISPR, statistical tests, etc.)
   • Results: sentences containing quantitative findings (p-values, fold changes, percentages)
   • Conclusions: closing statements or sentences with conclusion keywords

3. Entity Recognition

   • Gene/protein names (HGNC symbols, patterns like ALL_CAPS or CamelCase ≥3 chars)
   • Drug names and treatment modalities
   • Disease/condition names

4. Statistics Extraction

   • Extract p-values, confidence intervals, fold changes, effect sizes
   • Identify sample sizes (n=, N=)

5. Impact Assessment

   • Classify journal impact (high/medium/low based on known high-impact journals)
   • Identify publication type (original research, review, meta-analysis, clinical trial)

6. Generate Report

   • Write markdown with sections: Citation, Abstract, Key Findings, Methods, Results, Conclusions, Entities, Keywords, Impact Notes

Key Execution Patterns

Single paper digest:

python paper_digest_single.py --pmid 35769483 --outdir ./results

DOI with citations:

python paper_digest_single.py --doi 10.1038/s41586-022-04826-7 --include-citations 10 --outdir ./digest

arXiv preprint:

python paper_digest_single.py --arxiv 2301.12345 --outdir ./arxiv_digest

Local PDF (metadata from filename):

python paper_digest_single.py --pdf ./papers/my_study.pdf --outdir ./results

JSON output for programmatic use:

python paper_digest_single.py --pmid 35769483 --output-format json --outdir ./results

Topic-specific extraction:

python paper_digest_single.py --pmid 35769483 --topic-keywords "CRISPR,cancer,immunotherapy" --outdir ./results

Parameter Decision Guide

Parameter	When to Use	Example
--pmid	When you have a PubMed ID from pubmed.gov	--pmid 35769483
--doi	When citing a paper by DOI (most journals support this)	--doi 10.1038/nature12373
--arxiv	For preprints before peer review publication	--arxiv 2301.12345
--pdf	When you have a local copy and no ID available	--pdf ./paper.pdf
--include-citations	To include top cited references in output	--include-citations 5 for top 5
--topic-keywords	To highlight domain-specific findings	--topic-keywords "CRISPR,off-target"
--output-format	Use json for pipeline integration, markdown for reading	--output-format json

Failure Modes & Recovery

Error	Cause	Solution
"PMID not found"	Invalid PubMed ID or network issue	Verify PMID at pubmed.gov, check internet connection
"DOI resolution failed"	Invalid DOI format or CrossRef API timeout	Verify DOI format (10.xxxx/...), try using PMID instead
"arXiv ID not found"	Invalid ID format or not an arXiv paper	Verify ID at arxiv.org
"PDF parsing incomplete"	Complex PDF structure or scanned image	Try converting PDF to text first, or use PMID/DOI
"No abstract found"	Paper predates abstract requirement or restricted access	Use locally available metadata or title/keywords only
"Rate limit exceeded"	Too many API requests in short time	Wait 30 seconds before retrying
"Network timeout"	API server unreachable	Retry in 1-2 minutes, check network connectivity

Requirements

See requirements.txt for exact versions. Core dependencies:

• requests — HTTP API calls to PubMed, CrossRef, arXiv
• pandas — Data organization and export
• xml.etree.ElementTree — PubMed XML parsing (stdlib)
• json — Metadata serialization (stdlib)
• re — Regular expression entity extraction (stdlib)