AnyResearch

  日本語

AnyResearch is a MATLAB pipeline for keyword-based literature collection, research-trend review, and multi-institution benchmarking.
It uses OpenAlex and arXiv to fetch scholarly metadata and exports review-ready results to Excel.

Design philosophy: Convert publicly available scholarly metadata into structured, decision-ready material.
The user performs the final interpretation; AnyResearch focuses on delivering clean inputs for literature review and institutional comparison.

Typical Use Cases

Scenario	Question	What to use
Research theme selection	Which topics are growing fastest?	Layer 0: keyword search + Summary
Literature review	Collect review articles in a target field	Layer 0: filterType="review"
Competitive survey	What are rival institutions publishing?	Layer 1: institution batch
University IR / planning	Compare your institution against benchmark universities	Layer 1 + Layer 2
Grant proposal support	Add quantitative trend evidence	Layer 2: analytics
EasyMolKit integration	Find reproducible cheminformatics papers systematically	Layer 0: cheminformatics queries

Four-Layer Architecture

AnyResearch is designed as four incremental layers. Layer 0 alone covers the primary use case.

Layer	Additional requirements	What you get
Layer 0 (Core)	MATLAB + OpenAlex API Key	Keyword search, Excel workbook (4 sheets), JSONL / CSV output
Layer 1 (Batch)	institutions.csv	Multi-institution processing and batch_comparison.xlsx
Layer 2 (Analytics)	none	Citation velocity, topic growth rate, and institution dominance integrated into outputs
Layer 3 (PDF)	Text Analytics Toolbox or Python	OA PDF download, text extraction, keyword evidence

Quick Start

Standalone downstream examples built on search_results.jsonl are documented in docs/examples.md. They are optional and not part of the core product pipeline.

1. Set an OpenAlex API Key

1. Create a free account at openalex.org
2. Copy your API Key from openalex.org/settings/api
3. Create config/settings.json based on config/settings.example.json

{
  "openalex": {
    "api_key": "YOUR_API_KEY_HERE"
  }
}

2. Run a single keyword search

Edit Section 0 of main_run_pipeline.m:

query      = "renewable energy forecasting";
fromDate   = "2023-01-01";
toDate     = "2025-12-31";
sortBy     = "cited_by_count:desc";
filterType = "";

Search syntax:

• AND: separate with spaces
• OR: use |
• Phrase: wrap in quotes

Run Section 0 and then Section 1 with Run Section (Ctrl+Enter).
Outputs are saved under result/runs/<YYYYMMDD_HHMMSS>/.

3. Check outputs

result/runs/<YYYYMMDD_HHMMSS>/
  search_results.xlsx
  search_results.jsonl
  search_results.csv
  run_meta.json

Excel Output

Sheet	Contents
Overview	Title, DOI, publication year, citation count, OA flag, journal, abstract
Detail	Authors, affiliations, PDF status, topics, and other detailed fields
Summary	Yearly paper counts, average citations, citation velocity, growth rate
Config	Search conditions and run metadata

Optional Features

Layer 1: Benchmark-Institution Batch Mode

Process multiple universities or organizations in one run and generate cross-institution comparisons.

Step 1: Generate an institution candidate CSV

prepare_institutions_csv(["Example Research University", "Example Technical University", "Example Metropolitan University"], ...
    countryFilter="JP", maxCandidates=3)
% -> Outputs reviewed-v2 candidates to data/list/institutions_candidate.csv
% -> Review include / role / note and use it as institutions.csv

lookup_institution_id("Example Research University")

institutions_candidate.csv already uses the reviewed batch schema. No manual column renaming is required.
After reviewing include, save or reuse it as data/list/institutions.csv.

Step 2: Run the batch

main_run_batch

In main_run_batch.m:

• set prepareList=true to refresh the candidate CSV via Section 0.5
• set dryRun=true to preview filters and hit counts only
• run normally to write outputs under result/batch/<YYYYMMDD_HHMMSS>/

See also: Benchmark Institutions Workflow

Layer 2: Analytics

Analytics are integrated automatically into Summary and batch_comparison.xlsx.

Metric	Meaning
avg_citation_velocity	Average annual citation rate per paper
growth_rate_pct	Year-over-year paper-count growth
institution_dominance	Composite score based on paper share and citation share

Candidate Ledger (Phase L)

Candidate discovery can now be managed across runs instead of by one-off spreadsheet copies.

• appendToCandidates=true appends the final result table into result/candidates/candidates.jsonl
• candidates.xlsx provides a review view sorted for triage
• repro_candidates.md exports only status="reviewed" rows for EasyMolKit registration
• update_candidates_ledger(...) lets you mark rows as reviewed, rejected, or registered_RPxx

arXiv Integration

Fetch preprints from arXiv in parallel with OpenAlex:

useArxiv = true;

• Recorded as source_dataset = "arxiv"
• DOI duplicates are removed automatically
• arXiv preprints are excluded when filterType="article"

Layer 3: PDF Extension

Feature	Parameter	Description
PDF download and extraction	enablePdfDownload	Download OA PDFs and extract text
Keyword evidence	enableKeywordEvidence	Extract keyword hit snippets from PDF text

PDF extraction uses a two-stage engine:

• Engine 1: extractFileText() (Text Analytics Toolbox)
• Engine 2: Python pdfminer fallback

Directory Structure

main_run_pipeline.m
main_run_batch.m
src/
  openalex/
  adapters/
  export/
  pipeline/
  config/
  util/
  analytics/
  pdf/
  python/
config/
data/list/
result/
test/smoke/
docs/

Related Documents

File	Contents
docs/quickstart.md	Detailed setup, execution guide, and FAQ
docs/workflows/benchmark_institutions.md	Candidate generation, manual review, and batch execution
docs/workflows/repro_discovery.md	Reproduction-candidate discovery and candidate-ledger workflow
docs/reference.md	Function and smoke-test reference
CHANGELOG.md	High-level release history

Detailed internal project rules and phase-by-phase planning remain in the private development repository.

License

MIT License. See LICENSE for details.