literaturereport

KO | EN

gitlite — search

#javascript #hacktoberfest #llm #golang #deep-learning #macos #nodejs #awesome #cli #kubernetes #vue #openai #ios #mcp

literaturereport

(★ 17)

Automated skill for converting scientific papers into Chinese academic literature-report PowerPoint presentations, with figure extraction, reviewed panel cropping, structured paper interpretation, and speaker notes.

literaturereport Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# literaturereport

### 10 分钟把科研论文 PDF 变成中文文献汇报 PPT

![License](https://img.shields.io/badge/license-MIT-blue)
![Skill](https://img.shields.io/badge/Codex-skill-111827)
![Language](https://img.shields.io/badge/language-中文%20%7C%20English-green)
![Status](https://img.shields.io/badge/status-Beta-orange)
![Output](https://img.shields.io/badge/output-PPTX-purple)

[项目简介](#项目简介) · [PPT 展示](#ppt-展示) · [亮点](#亮点) · [技能索引](#技能索引) · [安装](#安装) · [使用流程](#使用流程) · [依赖配置](#依赖配置) · [仓库结构](#仓库结构) · [English](#english)

---

## 项目简介

还在为了组会文献汇报发愁吗？

读一篇文献要 3 天，做 PPT 又要 3 小时；图要一张张截，panel 要一个个拆，排版要反复调，最后还得给每页写汇报批注。等 PPT 做完，人已经比论文里的 negative control 还沉默。

`literaturereport` 就是为这种时刻准备的文献汇报 skill：给它一篇论文 PDF，它会帮你在短时间内整理出一套可用于组会、读书报告或论文分享的中文学术汇报 PPT。它不仅负责生成页面，还会补充讲稿提示、整理 figure 解读、归纳总结与应用启发，让你上台时不至于只会说“这个图说明了一个现象”。最后的总结与应用部分也会尽量把论文价值讲清楚，争取让导师从“你再想想”切换到“这个总结不错”。

（作者声明：不支持任何利用人工智能途径糊弄导师及同门。该 skill 本着将广大研究生从繁杂的 PPT 制作中解脱出来的理念，希望用户把更多时间投入到文献阅读以及批判性理解中。）

一句话：

> 把论文 PDF 变成一套能讲、能看、能继续修改的中文文献汇报 PPT。

---

## PPT 展示

仓库内提供一份完整的 22 页示例 PPT，用于展示 `literaturereport` 对研究思路、论文 Figure、结论讨论和批判性阅读页面的实际组织效果。

![literaturereport 示例 PPT：研究思路、研究结果、结论与讨论、不足与优化](https://raw.githubusercontent.com/doczrz/literaturereport/main/assets/showcase/literaturereport-showcase.png)

预览页依次展示：**研究思路 → 代表性研究结果 → 结论与讨论 → 不足与优化**。

[查看或下载完整示例 PPTX](./Chen-2025-single-cell-spatial-transcriptomics-literature-report-limitations-closing-corrected.pptx)

> 示例文件用于展示生成效果。论文图片及论文相关内容的权利归原作者和出版机构所有。

---

## 亮点

| 能力 | 说明 |
|---|---|
| 论文结构化阅读 | 提取题目、摘要、研究背景、研究对象、研究思路、主要结果、讨论和结论 |
| figure 自动提取 | 使用 `pdffigures2` 提取候选 figure，并生成需要用户确认的图片材料 |
| figure 人工确认 | 支持用户审核和修正 figure crop，减少坐标轴、图例、比例尺、panel label 被截断的问题 |
| panel 级别解读 | 结合 figure legend 和 Results 解读 figure，而不是只看图片或只抄图注 |
| PPT 自动生成 | 生成可编辑中文学术汇报 PPTX，并为每页添加 speaker notes |
| 中文学术风格 | 默认使用脚本化中文学术版式，不依赖私人 PPT 模板 |

---

## 技能索引

| Skill | 状态 | 用途 | 触发词 |
|---|---|---|---|
| [`literaturereport`](SKILL.md) | Beta | 从科研论文 PDF 生成中文文献汇报 PPTX，包含 figure 解读、汇报结构和 speaker notes | “文献汇报 PPT”、 “paper to PPT”、 “journal club slides”、 “组会汇报” |

---

## 适合谁

- 需要做组会文献汇报的研究生。
- 需要快速准备 journal club 的科研人员。
- 想把论文从“读过了”变成“讲得清楚”的用户。
- 不想把晚上都花在截图、排版和写备注上的人。

---

## 安装

### 方式一：作为 Codex skill 安装

推荐把完整仓库 clone 到本地，然后复制整个 `literaturereport` 目录到 Codex skills 目录。

```bash
git clone https://github.com/doczrz/literaturereport.git
```

Windows 常见目标路径：

```text
C:\Users\<你的用户名>\.codex\skills\literaturereport
```

macOS / Linux 常见目标路径：

```text
~/.codex/skills/literaturereport
```

关键规则：保留完整目录结构。不要只复制 `SKILL.md`，因为脚本、assets 和依赖说明也会被用到。

### 方式二：手动下载

如果你从 GitHub 网页下载 zip，解压后请确认目录结构类似：

```text
literaturereport/
├── SKILL.md
├── README.md
├── scripts/
├── assets/
└── agents/
```

然后把整个 `literaturereport/` 文件夹放入 Codex skills 目录。

---

## 使用流程

安装后，开启新的 Codex 会话，自然描述任务即可，例如：

```text
使用 $literaturereport，把 C:\path\to\paper.pdf 做成中文文献汇报 PPT。
```

如果希望沿用已有 PPT 的字体和配色，可在同一条请求中附上主题 PPTX；它只作为视觉风格来源，不会覆盖论文证据或汇报结构。

一次完整运行通常包括：

1. **解析论文**：检查本地英文 PDF 是否包含可选择文本，并提取标题、摘要、章节、图注和正文证据。
2. **建立证据包**：先生成结构化 evidence packet，再组织研究背景、研究思路和结果页，避免直接从未解析的 PDF 凭印象写内容。
3. **处理 Figure**：按图注建立主 Figure 清单，执行一次 `pdffigures2`；缺失或不完整的 Figure 会从原始高分辨率页面恢复。
4. **安全处理 panel**：密集 Figure 的 panel 标签坐标会一次性批量确认；没有可靠分割边界时保留完整 Figure，不强行裁切。
5. **生成汇报结构**：根据原始研究或叙述性综述选择对应叙事，完成中文页面文案和可直接阅读的 speaker notes。
6. **编译与检查**：编译前验证内容和证据锚点，编译后检查全部渲染页、布局、溢出、Figure 完整性和讲稿。
7. **交付成品**：只有最终 QA 通过后才交付可编辑 `.pptx`。

当论文缺少可恢复的 Figure，或需要确认原文是否包含研究流程图时，skill 会把问题合并后一次性询问。通常用户只需提供 PDF，并在必要时完成这一轮确认。

---

## 依赖配置

需要：

- Python 3.10+
- Pillow
- python-pptx
- Java runtime
- 本地配置的 `pdffigures2.jar`

本仓库不内置 `pdffigures2.jar`。它体积较大，也不适合作为源码文件直接提交。

不要提交以下文件：

```text
assets/dependencies/pdffigures2.jar
assets/dependencies/figure_dependencies.json
```

其中 `figure_dependencies.json` 会记录本机绝对路径，应由每台机器首次使用时自动生成。

如果本机已经安装 Java，并且已有 `pdffigures2.jar`：

```bash
python scripts/install_figure_dependencies.py
```

如果 Java 或 jar 文件在自定义位置：

```bash
python scripts/install_figure_dependencies.py --java "C:\path\to\java.exe" --pdffigures2-jar "C:\path\to\pdffigures2.jar"
```

如果本机没有 jar，可以在用户确认下载来源后提供 URL：

```bash
python scripts/install_figure_dependencies.py --pdffigures2-url "https://user-confirmed-url/pdffigures2.jar"
```

配置完成后会生成：

```text
assets/dependencies/figure_dependencies.json
```

后续 figure 提取脚本会自动读取该配置。

---

## 输出内容

一次完整运行通常会产生：

- 可编辑的 `.pptx` 文献汇报文件
- 候选 figure 和已审核 figure crop
- figure contact sheet
- 结构化内容 JSON
- 每页 PPT 的 speaker notes

这些运行产物属于具体论文项目输出，不应提交到 skill 仓库。

---

## 仓库结构

```text
literaturereport/
├── SKILL.md
├── README.md
├── RELEASE_AUDIT.md
├── agents/
├── assets/
│   ├── qi-2022-v4-reference-structure.json
│   └── dependencies/
│       └── THIRD_PARTY_NOTICES.md
└── scripts/
    ├── generate_literature_report_deck.py
    ├── extract_figures_pdffigures2.py
    ├── validate_reviewed_crops.py
    ├── approve_figure_review.py
    └── test_*.py
```

---

## 测试

```bash
python scripts/test_extract_figures_selection.py
python scripts/test_panel_split_validation.py
```

测试重点包括：

- figure 选择逻辑是否稳定
- panel 分割是否避免明显裁切
- 本机依赖配置是否可被脚本读取
- 生成 PPT 时是否残留内部流程词或测试材料

---

## 注意事项

- `pdffigures2.jar` 需要用户本机配置，仓库只保留安装脚本和第三方说明。
- 生成的 PPTX、论文图片、contact sheet、论文文本提取文件都不属于仓库源码。
- 默认主题是脚本生成的中文学术风格，不需要私有模板。
- 如果使用本地模板，应确保用户有权使用该模板。
- 该 skill 更适合“可讲的文献汇报”，不是全文翻译工具，也不是论文排版工具。

---

## English

### Turn scientific paper PDFs into Chinese journal-club PowerPoint decks

[Overview](#overview) · [Highlights](#highlights) · [Skill Index](#skill-index) · [Example Output](#example-output) · [Install](#install-1) · [Workflow](#workflow) · [Dependencies](#dependencies) · [License](#license)

## Overview

Still losing sleep over journal club slides?

You spend three days reading a paper, another three hours building the PowerPoint, then somehow still need to crop every figure, split every panel, fix the layout, and write speaker notes. By the time the deck is done, you understand the formatting pain better than the paper itself.

`literaturereport` is a Codex / ChatGPT skill built for exactly that situation. Give it a scientific paper PDF, and it helps turn the paper into a polished Chinese academic literature-report PowerPoint deck for group meetings, journal clubs, seminars, or paper sharing. It also prepares figure interpretation, speaker notes, and a final summary/application section, so your advisor has a better reason to say, "This is actually a useful takeaway."

Author's note: this project does not support using AI to bluff your advisor or labmates. The goal is to free researchers from repetitive slide-making work, so more time can go into reading papers carefully and thinking critically.

In short:

> Turn a paper PDF into an editable, presentation-ready Chinese academic PPT deck with figure interpretation and speaker notes.

## Highlights

| Capability | Description |
|---|---|
| Structured paper reading | Extracts title, abstract, background, research object, study logic, results, discussion, and conclusions |
| Figure extraction | Extracts candidate figures with `pdffigures2` |
| Human figure review | Supports reviewed crops to avoid cutting axes, legends, scale bars, and panel labels |
| Panel-level interpretation | Reads figures together with figure legends and Results text |
| PPTX generation | Generates editable Chinese academic PPTX decks with speaker notes |
| Academic styling | Uses a scripted academic theme by default |

## Skill Index

| Skill | Status | Purpose | Trigger words |
|---|---|---|---|
| [`literaturereport`](SKILL.md) | Beta | Generate Chinese academic literature-report PPTX decks from scientific PDFs | “paper to PPT”, “journal club slides”, “literature report”, “组会汇报” |

## Example Output

The repository includes a complete 22-slide example deck showing how `literaturereport` organizes the study design, paper Figures, conclusion, and critical appraisal.

![literaturereport example deck: study design, results, conclusion, and limitations](https://raw.githubusercontent.com/doczrz/literaturereport/main/assets/showcase/literaturereport-showcase.png)

The four preview pages show: **study design → representative result → conclusion and discussion → limitations and optimization**.

[View or download the complete example PPTX](./Chen-2025-single-cell-spatial-transcriptomics-literature-report-limitations-closing-corrected.pptx)

> The example is provided to demonstrate the generated presentation. Rights to the paper figures and paper-specific content remain with the original authors and publisher.

## Install

Clone the repository:

```bash
git clone https://github.com/doczrz/literaturereport.git
```

Copy the full `literaturereport/` folder into your Codex skills directory. Keep the full folder structure; do not copy only `SKILL.md`.

## Workflow

After installation, start a new Codex session and ask naturally:

```text
Use $literaturereport to turn C:\path\to\paper.pdf into a Chinese literature-report PowerPoint deck.
```

An existing PPTX may be supplied in the same request as a visual-theme source. It can contribute fonts and colors, but it does not replace the paper evidence or the presentation structure.

A complete run normally follows these stages:

1. **Parse the paper**: verify that the local English PDF contains selectable text and extract its title, abstract, sections, legends, and source passages.
2. **Build the evidence packet**: create structured evidence before writing the background, study-design, and result slides.
3. **Resolve Figures**: inventory every main Figure from the legends, run `pdffigures2` once, and recover missing or rejected Figures from high-resolution page renders.
4. **Handle panels safely**: confirm dense-Figure panel coordinates in one batch; preserve the whole Figure whenever no reliable split boundary exists.
5. **Author the deck**: select the original-research or narrative-review structure and write Chinese slide copy plus directly readable speaker notes.
6. **Compile and validate**: check the content and evidence anchors before compilation, then inspect all rendered slides, layouts, overflow, Figures, and notes.
7. **Deliver the PPTX**: return the editable deck only after the final QA report passes.

If a Figure cannot be recovered automatically, or the presence of an original workflow diagram must be confirmed, the skill combines those questions into one user handoff. In most cases, the user only supplies the PDF and completes that single confirmation when needed.

## Dependencies

- Python 3.10+
- Pillow
- python-pptx
- Java runtime
- A local `pdffigures2.jar`, configured on first use

`pdffigures2.jar` is not bundled in this repository. Do not commit:

```text
assets/dependencies/pdffigures2.jar
assets/dependencies/figure_dependencies.json
```

## License

MIT

literaturereport

literaturereport Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

literaturereport

literaturereport Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README