Quant Paper Hub is an AI-assisted quantitative research operations platform for discovering, organizing, reading, summarizing, and converting research papers into actionable quantitative research ideas.
The project focuses on quantitative finance, machine learning for asset pricing, stock return prediction, factor investing, portfolio optimization, market regime analysis, and China A-share research.
It is not an investment advisory system, not a stock recommendation system, and not an automated trading system.
Current documentation version:
V0.9.7 Public Release Candidate / Pre-V1 Final Check
Current stable feature version:
V0.9.4 User Manual and UI Operation Guide
Quant Paper Hub is designed to support a full research workflow:
paper discovery
-> abstract AI summary
-> PDF download
-> full-text extraction
-> chunk-based full-paper summary
-> reading notes and vocabulary cards
-> research ideas
-> experiment tasks
-> reading card export
-> research dashboard
-> Semantic Scholar citation network
-> future literature review generation
The long-term goal is to build a personal research intelligence layer for quantitative finance research and A-share factor strategy development.
Current capabilities include:
- arXiv keyword-based paper discovery
- paper metadata storage in SQLite
- Streamlit-based Paper Reader
- reading status, favorites, manual tags, and relevance score
- vocabulary annotations
- DeepSeek AI abstract summaries
- PDF download and full-text extraction
- quick full-text summary
- chunk-based full-paper summarization
- reading notes
- research idea board
- experiment task backlog
- reading card export
- reading card library
- research dashboard
- daily update pipeline
- maintenance panel
- Semantic Scholar metadata integration
- citation count, reference count, and influential citation count
- Semantic Scholar references and forward citations
- Semantic Scholar citation network graph
- OpenAlex metadata integration
- DOI lookup and OpenAlex cited_by_count
- OpenAlex open-access status and PDF URL discovery
- Unpaywall open-access lookup
- unified open-access PDF resolver
- resolved_pdf_url and resolved_pdf_source fields
- user manual for UI controls and maintenance workflows
The main paper reading interface. It supports paper filtering, citation metrics, abstract viewing, AI summaries, vocabulary cards, notes, research ideas, full-text summary, Semantic Scholar network, and reading card export.
A global board for research ideas extracted from papers or manually created during reading.
A backlog for converting research ideas into future experiments.
A library of exported Markdown reading cards.
Dashboard for paper, summary, idea, and reading progress metrics.
Maintenance utilities for daily pipeline, reports, backups, and project health checks.
The V0.8 feature line adds Semantic Scholar integration.
Added Semantic Scholar metadata foundation:
- metadata schema
- API client
- fetch script
- coverage report
- arXiv ID matching
- title fallback search
Integrated citation metrics into Paper Reader:
- citation_count
- reference_count
- influential_citation_count
- venue
- year
- match method
- match confidence
- citation filters
- Semantic Scholar metadata card
Added Semantic Scholar references and citations foundation:
- semantic_scholar_references table
- semantic_scholar_citations table
- references endpoint fetch
- citations endpoint fetch
- edge coverage report
Added Semantic Scholar Network UI:
- S2 Network tab
- citation network graph
- backward references table
- forward citations table
- Has S2 Network filter
- network edge counts in the paper table
In the S2 Network tab:
- Center node: selected paper
- Right side / Backward References: papers cited by the selected paper
- Left side / Forward Citations: papers that cite the selected paper
Important limitation:
The S2 Network is an external citation network. It is not yet a semantic relatedness model. A paper that cites the selected work may come from another domain, especially for AI, ML, RL, or LLM-related papers.
The V0.9 feature line adds OpenAlex, Unpaywall, and unified PDF discovery.
Added OpenAlex metadata integration foundation:
- OpenAlex API client
- OpenAlex metadata database table
- DOI matching
- inferred arXiv DOI matching
- title fallback search
- OpenAlex coverage report
Integrated OpenAlex metadata into Paper Reader:
- DOI
- OpenAlex Work ID
- OpenAlex cited_by_count
- OpenAlex open-access status
- OpenAlex OA URL
- OpenAlex best OA PDF URL
- OpenAlex filters
- OpenAlex metadata card
Added Unpaywall open-access lookup foundation:
- Unpaywall API client
- Unpaywall metadata database table
- DOI-based OA lookup
- OA status
- repository copy detection
- best OA location
- best OA PDF URL when available
Added open-access PDF discovery resolver:
- unified
resolved_pdf_url - unified
resolved_pdf_source - direct PDF source priority across original paper source, OpenAlex, and Unpaywall
- separate OA landing-page resolution
- Full Text tab now prioritizes
resolved_pdf_url
PDF resolver priority:
- original
papers.pdf_url - OpenAlex
best_oa_pdf_url - Unpaywall
best_oa_pdf_url
OA landing-page priority:
- Unpaywall
best_oa_url - OpenAlex
oa_url
Landing pages are shown for manual access and are not treated as direct PDF download targets.
Added user manual for UI controls, Maintenance pipeline options, PDF discovery logic, and safe operation workflows.
See:
quant_paper_hub/
config/
keywords.yaml
data/
papers.db
docs/
PROJECT_REPORT_*.md
REFACTOR_REPORT_V0.7.9.md
SEMANTIC_SCHOLAR_INTEGRATION_REPORT_V0.8.4.md
USER_MANUAL_V0.9.4.md
USAGE.md
reports/
weekly_quant_papers_*.md
semantic_scholar_metadata_fetch_*.md
semantic_scholar_edges_fetch_*.md
openalex_metadata_fetch_*.md
unpaywall_metadata_fetch_*.md
scripts/
01_fetch_arxiv.py
02_generate_weekly_digest.py
07_daily_update.py
18_fetch_semantic_scholar_metadata.py
19_fetch_semantic_scholar_edges.py
20_fetch_openalex_metadata.py
21_fetch_unpaywall_metadata.py
src/
arxiv_client.py
db.py
semantic_scholar_client.py
semantic_scholar_db.py
semantic_scholar_network.py
openalex_client.py
openalex_db.py
unpaywall_client.py
unpaywall_db.py
pdf_utils.py
web/
streamlit_app.py
app_pages/
components/
README.md
README.zh-CN.md
requirements.txt
Create and activate the virtual environment:
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
Run the Streamlit app:
streamlit run web/streamlit_app.py --server.address 0.0.0.0 --server.port 8501
Fetch arXiv papers:
python scripts/01_fetch_arxiv.py
Generate weekly digest:
python scripts/02_generate_weekly_digest.py
Fetch Semantic Scholar metadata:
python scripts/18_fetch_semantic_scholar_metadata.py --limit 30 --sleep 1.5
Fetch Semantic Scholar references and citations:
python scripts/19_fetch_semantic_scholar_edges.py --limit-papers 10 --reference-limit 20 --citation-limit 20 --sleep 1.5
Fetch OpenAlex metadata:
python scripts/20_fetch_openalex_metadata.py --limit 30 --sleep 1.5
Fetch Unpaywall metadata:
export UNPAYWALL_EMAIL="your_email@example.com"
python scripts/21_fetch_unpaywall_metadata.py --limit 30 --sleep 1.5
Current sources:
- arXiv
- Semantic Scholar
- OpenAlex
- Unpaywall
Planned sources:
- Open-access publisher metadata
- domain-specific related-paper ranking
- embedding-based literature similarity
The default keyword configuration focuses on:
- machine learning asset pricing
- stock return prediction
- factor investing
- portfolio optimization
- financial time series
- cross-sectional stock return prediction
- distribution shift in stock markets
- market regime detection
- walk-forward validation in finance
- China A-share factor research
OpenAlex metadata integration foundation.
OpenAlex metadata integrated into Paper Reader UI.
Unpaywall open-access lookup foundation.
Open-access PDF discovery resolver.
User manual for UI controls, Maintenance pipeline options, PDF discovery logic, and safe operation workflows.
README and documentation synchronization for the V0.9.x metadata and PDF discovery line.
Open-source readiness checklist, .gitignore hardening, and secret-safety documentation.
Public release candidate and Pre-V1 final check.
Complete research review generation workflow.
Planned V1.0 workflow:
paper discovery
-> metadata enrichment
-> PDF discovery
-> full-text extraction
-> structured summary
-> citation network analysis
-> research idea generation
-> experiment task generation
-> literature review draft
Quant Paper Hub is designed as the research intelligence layer for the A-share multi-factor candidate pool and risk filtering system.
Quant Paper Hub discovers, organizes, and interprets research papers.
The A-share quant system converts useful research ideas into factor engineering, model validation, walk-forward experiments, risk filtering logic, and performance analysis.
This project is for research assistance, academic reading, and quantitative research workflow management only.
It does not provide investment advice, trading signals, stock recommendations, or guaranteed returns.
Generated data, reports, databases, logs, PDF files, API keys, environment files, and local caches should not be committed to Git.