| |
| """Rebuild both tracks and bilingual entry points without changing admissions.""" |
| import csv |
| import json |
| import os |
| import subprocess |
| import sys |
| from collections import Counter |
| from pathlib import Path |
|
|
| ROOT=Path(__file__).resolve().parents[1] |
| VENUES=('ICML','ICLR','ACL','EMNLP','NeurIPS','AAAI') |
|
|
|
|
| def read(track): |
| with (ROOT/track/'finance_catalog.csv').open(encoding='utf-8-sig',newline='') as stream: |
| return list(csv.DictReader(stream)) |
|
|
|
|
| def main(): |
| rows={track:read(track) for track in ('main','findings')} |
| schema=json.loads((ROOT/'main/checks/finance_taxonomy/schema.json').read_text()) |
| for track in rows: |
| env={**os.environ,'AI4FINANCE_TRACK':track} |
| for script in ('export_collection.py','build_indexes.py'): |
| subprocess.run([sys.executable,str(ROOT/'scripts'/script)],env=env,check=True) |
| for lang in ('en','zh'): |
| title='Main' if track=='main' else 'Findings' |
| data=rows[track];counts=Counter(r['finance_scope'] for r in data) |
| pdfs=sum(r['download_status']=='downloaded' for r in data) |
| links=('[English](README.md) | [中文](README.zh-CN.md) | [Repository home](../README.md)' if lang=='en' else |
| '[English](README.md) | [中文](README.zh-CN.md) | [仓库首页](../README.zh-CN.md)') |
| intro=(f'{len(data):,} long papers: {counts["core"]:,} finance core and {counts["application"]:,} financial applications; {pdfs:,} local PDFs.' if lang=='en' else |
| f'共 {len(data):,} 篇长文：金融核心研究 {counts["core"]:,} 篇、金融应用 {counts["application"]:,} 篇；本地 PDF {pdfs:,} 份。') |
| lines=[f'# AI for Finance — {title}', '', links, '', intro, '', |
| ('This directory contains only '+title+' long papers. Local paths in its CSV/JSONL/Excel are relative to this directory.' if lang=='en' else |
| '本目录仅存储 '+title+' 长文。CSV、JSONL 与 Excel 中的本地路径均相对于本目录。'), '', |
| ('## Browse' if lang=='en' else '## 浏览'), '', |
| '- [All papers / 全部论文](indexes/all_papers.md)', |
| '- [Finance tasks / 金融任务](indexes/ai_for_finance.md) · [Subtopics / 子方向](indexes/subtopics.md) · [Methods / 技术方法](indexes/methods.md)', |
| '- [Core / 金融核心研究](indexes/core.md) · [Applications / 金融应用](indexes/applications.md)', |
| '- [Excel](finance_catalog.xlsx) · [CSV](finance_catalog.csv) · [JSONL](finance_catalog.jsonl) · [BibTeX](finance_references.bib)', |
| '- [Taxonomy / 分类说明](indexes/finance_taxonomy.md) · [Selection evidence / 筛选依据](checks/README.md)', '', |
| '| Venue / 会议 | Long papers / 长文 |', '|---|---:|'] |
| for venue in VENUES: |
| subset=[r for r in data if r['venue']==venue] |
| if subset:lines.append(f'| [{venue}](indexes/venues/{venue.lower()}.md) | {len(subset)} |') |
| lines+=['', ' · '.join(f'[{year}](indexes/years/{year}.md)' for year in sorted({r['year'] for r in data},reverse=True)), '', |
| ('Papers without verified long-paper eligibility are kept out of this catalog. See the [screening audit](../checks/track_screening/README.md) for coverage and pending cases.' if lang=='en' else |
| '无法核验长文属性的论文不进入正式清单；筛查覆盖和待核验记录见[筛查审计](../checks/track_screening/README.md)。'), ''] |
| lines[4:4]=['**Long papers only. Short papers are excluded. / 仅收录长文，不收录短文。**', ''] |
| (ROOT/track/('README.md' if lang=='en' else 'README.zh-CN.md')).write_text('\n'.join(lines)) |
| venue_table=['| Venue / 会议 | Main | Findings |','|---|---:|---:|'] |
| for venue in VENUES: |
| cells=[] |
| for track in rows: |
| count=sum(r['venue']==venue for r in rows[track]) |
| cells.append(f'[{count}]({track}/indexes/venues/{venue.lower()}.md)' if count else ('—' if track=='findings' and venue not in ('ACL','EMNLP') else '0')) |
| venue_table.append(f'| {venue} | '+ ' | '.join(cells)+' |') |
| topic_table=['| Finance task / 金融任务 | Main | Findings |','|---|---:|---:|'] |
| for topic in schema['topics']: |
| cells=[f"[{sum(topic['id'] in r['finance_topic_ids'].split('; ') for r in rows[track])}]({track}/indexes/finance_topics/{topic['id']}.md)" for track in rows] |
| topic_table.append(f"| {topic['label']} | "+' | '.join(cells)+' |') |
| for lang in ('en','zh'): |
| zh=lang=='zh' |
| lines=['# Awesome AI4Finance Papers', '', '[English](README.md) | [中文](README.zh-CN.md)', '', |
| ('按金融任务、子方向和技术方法整理的 AI for Finance 长文论文库，由 [OpenFinMind](https://github.com/OpenFinMind) 维护。' if zh else |
| 'AI for Finance long papers organized by financial task, subtopic, and technical method. Maintained by [OpenFinMind](https://github.com/OpenFinMind).'), '', |
| '## '+('收录范围' if zh else 'Scope'), '', |
| ('收录 **2017–2026** 年 **ICML、ICLR、ACL、EMNLP、NeurIPS、AAAI Main** 及 **ACL、EMNLP Findings** 中与金融相关的长文，涵盖金融核心研究与金融应用。Main 与 Findings 分开存储，并按金融任务、子方向和技术方法整理。' if zh else |
| 'Covers finance-related long papers from **ICML, ICLR, ACL, EMNLP, NeurIPS, and AAAI Main**, and **ACL / EMNLP Findings**, within **2017–2026**, including finance core research and financial applications. Main and Findings are stored separately and organized by financial task, subtopic, and technical method.'), '', |
| '## '+('独立目录' if zh else 'Separate collections'), '', |
| '| '+('类别 | 长文 | 论文与分类 | 数据导出 |' if zh else 'Track | Long papers | Papers and classifications | Exports |'), |
| '|---|---:|---|---|'] |
| for track in rows: |
| n=len(rows[track]);title=track.title() |
| lines.append(f'| [{title}]({track}/README'+('.zh-CN.md' if zh else '.md')+f') | {n:,} | [Tasks]({track}/indexes/ai_for_finance.md) · [Subtopics]({track}/indexes/subtopics.md) · [Methods]({track}/indexes/methods.md) | [Excel]({track}/finance_catalog.xlsx) · [CSV]({track}/finance_catalog.csv) · [JSONL]({track}/finance_catalog.jsonl) · [BibTeX]({track}/finance_references.bib) |') |
| lines+=['', ('Main 与 Findings 各有独立目录、索引、逐篇笔记、CSV、JSONL、Excel、BibTeX 和 PDF。导出文件中的 `local_path` 相对于所属类别目录；根目录不提供混合两类的论文清单。' if zh else |
| 'Main and Findings each have their own indexes, paper notes, CSV, JSONL, Excel, BibTeX, and PDFs. Exported `local_path` values are relative to the corresponding track directory.'), '', |
| '## '+('按会议浏览' if zh else 'By venue'), '', '\n'.join(venue_table), '', |
| '## '+('按金融任务浏览' if zh else 'By financial task'), '', |
| ('金融任务允许多标签，计数存在重叠；每篇另有唯一主方向。子方向描述具体任务，技术方法独立标注 LLM、强化学习、图学习、扩散、因果等，避免把技术名称当成金融任务。' if zh else |
| 'Financial tasks allow multiple labels, so counts overlap; every paper also has one primary task. Subtopics describe more specific problems. Technical methods such as LLMs, reinforcement learning, graph learning, diffusion, and causal methods are recorded separately.'), '', |
| '\n'.join(topic_table), '', '## '+('分类说明' if zh else 'Classification'), '', |
| ('- **金融核心研究（core）**：研究问题直接面向金融。\n- **金融应用（application）**：通用方法含有实质性的金融数据、实验、业务场景、建模或领域综述；不等同于金融专用方法。' if zh else |
| '- **Core:** the research problem directly concerns finance.\n- **Application:** a general method has substantive financial data, experiments, business cases, modeling, or a domain review; it need not be a finance-specific method.'), '', |
| ('筛选与核验说明见[筛查审计](checks/track_screening/README.md)。' if zh else |
| 'Selection and verification details are available in the [screening audit](checks/track_screening/README.md).'), '', |
| '## '+('资源' if zh else 'Resources'), '', |
| ('PDF 位于 `<main|findings>/pdfs/<venue>/<year>/`；官方版本与作者 arXiv 版本分别标注，并记录 SHA-256。未下载的已收录论文保留在线链接。论文版权归作者与出版方。维护方式见 [CONTRIBUTING.md](CONTRIBUTING.md)。' if zh else |
| 'PDFs live under `<main|findings>/pdfs/<venue>/<year>/`. Publisher and author arXiv versions are distinguished and hashed with SHA-256. Admitted papers without local PDFs retain online links. Paper copyrights remain with the authors and publishers. See [CONTRIBUTING.md](CONTRIBUTING.md) for maintenance.'), ''] |
| lines[4:4]=['[GitHub](https://github.com/OpenFinMind/Awesome-AI4Finance-Papers) · [Hugging Face](https://huggingface.co/datasets/OpenFinMind/Awesome-AI4Finance-Papers)', '', |
| '**仅收录长文，不收录短文。**' if zh else '**Long papers only. Short papers are excluded.**', ''] |
| (ROOT/('README.zh-CN.md' if zh else 'README.md')).write_text('\n'.join(lines)) |
|
|
|
|
| if __name__=='__main__':main() |
|
|
