Research Hunter

Research Hunter automates the tedious first step of a literature review: finding papers, filtering noise, and prioritizing what to read next.

Why use this?

When you search manually, you tend to:

• run the same queries repeatedly across sessions
• lose track of what you already skimmed
• get overwhelmed by volume and miss the highest-impact items

Research Hunter helps by producing a ranked, exportable shortlist so you can spend time reading—not wrangling search results.

What does it search?

Right now, Research Hunter uses the Semantic Scholar Graph API, which indexes millions of papers across disciplines.

What is the effect?

Research Hunter doesn’t just retrieve papers—it scores and ranks them using a transparent 0–100 rubric:

• Relevance (keyword matches)
• Impact (citations)
• Recency (publication year)

Outputs are written to:

• outputs/results.csv
• outputs/results.json

Install

Requires Python 3.10+.

python -m venv .venv
source .venv/bin/activate
pip install -U pip
pip install -e .

Configuration

Create a .env file (see .env.example).

Scoring is configurable for offline runs via:

• configs/scoring_config.example.json

Optional but recommended:

• SEMANTIC_SCHOLAR_API_KEY to increase rate limits

If you hit HTTP 429 (rate limit), add the key and retry.

More details: docs/api_setup.md.

Quick comparison

Feature	Standard search engines	Research Hunter
Transparent, adjustable scoring	❌	✅ (documented + configurable)
Relevance + impact + recency ranking	Mixed / implicit	✅ (explicit 0–100 rubric)
Export clean CSV/JSON	Sometimes	✅
Offline corpus analysis (no API calls)	❌	✅ (analyze mode)
Reproducible runs (config-driven)	❌	✅

Pipeline (high level)

flowchart LR
  A[Data Input\n(online search or local JSON)] --> B[Scoring Engine\nRelevance + Impact + Recency]
  B --> C[Ranked Output\nCSV + JSON]

Loading

Usage

Search for papers and export results:

research-hunter search \
  --query "emotion regulation ecological momentary assessment" \
  --limit 25 \
  --outdir outputs

Offline scoring of a local corpus (recommended for reproducible workflows):

research-hunter analyze \
  --input samples/sample_corpus.json \
  --output outputs/results.csv \
  --config configs/scoring_config.example.json

Output format

See samples/ for example CSV/JSON outputs.

Docs

• docs/architecture.md
• docs/scoring_methodology.md
• docs/api_setup.md

How to cite

A CITATION.cff is provided for GitHub/Zenodo-friendly citation metadata.

If you use Research Hunter in academic work, cite the software repository (and include the version/tag you used).

Contributing

Issues and PRs are welcome. Keep changes:

• reproducible
• transparent (document scoring changes)
• free of secrets (never commit .env)

License

MIT (see LICENSE).