Overview
This skill uses the Qinyan Academic OpenAPI to search a large number of related literature and, combined with AI analysis capabilities, provides researchers with systematic topic selection suggestions. The analysis covers topic directions, research value, research content overview, research difficulty analysis, and related literature recommendations.
Academic Integrity Statement: All topic suggestions are analyzed based on actually retrieved literature, and no literature or research trends are fabricated. The suggestions are for reference only, and the final topic decision should be combined with the advisor's opinion and personal research interests.
Prerequisites
Before use, please ensure:
- Set
QINYAN_API_KEYenvironment variable (go to https://platform.qinyanai.com/ to apply) - Available in the environment
curlandpython3
export QINYAN_API_KEY="your-api-key-here"
Workflow
Step 1: Search a large number of related literature
Based on the research field/direction provided by the user, use scripts/search.sh to search a large number of literature from multiple databases.
Search strategy:
- Chinese field: prioritize Wanfang + Google Scholar
- English field: prioritize Google Scholar + ArXiv/PubMed
- Medical: must search PubMed (Chinese query needs to be translated into English)
- Science, engineering, and computer science: must search ArXiv (Chinese query needs to be translated into English)
- Search at least 50-100 literature items from each relevant database to ensure sufficient coverage
# 检索示例
bash scripts/search.sh google '{"query": "large language models education", "max_results": 20, "date_from": "2022"}'
bash scripts/search.sh arxiv '{"query": "large language models education", "max_results": 50, "date_from": "2022"}'
bash scripts/search.sh wanfang '{"query": "大语言模型 教育应用", "max_results": 50, "date_from": "2022"}'
The search script parameter format is the same as the literature search skill:
| Data source | Command | Special notes |
|---|---|---|
| Google Scholar | bash scripts/search.sh google '<json>' |
Maximum 20 items per search |
| Wanfang | bash scripts/search.sh wanfang '<json>' |
Supports Chinese, 1-100 items |
| PubMed | bash scripts/search.sh pubmed '<json>' |
English only, 1-200 items |
| ArXiv | bash scripts/search.sh arxiv '<json>' |
English only, 1-100 items |
Step 2: Literature analysis and topic generation
Based on the large number of retrieved literature, analyze the research status, hot trends, and research gaps, and generate topic suggestions.
Step 3: Output structured topic suggestions
Output format
For each recommended topic, output according to the following structure:
Topic suggestion template
## 选题方向 [序号]：[选题名称]
### 选题方向
[具体的研究方向和切入点描述]
### 研究价值
[该选题的理论价值和实践意义，为什么值得研究]
### 研究内容概述
[该选题需要研究的主要内容、技术路线或方法框架]
### 研究难点分析
[该选题可能面临的主要技术难点、数据获取困难、方法挑战等]
### 相关文献
[列出5-10篇与该选题最相关的文献，标注来源]
- [1] 作者. 标题. 期刊/会议, 年份. DOI/URL
- [2] ...
Analysis requirements
- Number of topics: Based on literature coverage, recommend 3-5 differentiated topic directions
- Topic quality：
- Avoid overly broad topics (such as "deep learning research")
- Avoid overly narrow topics (such as "improvement of a certain metric on a specific dataset")
- Topics should be innovative and fill gaps in existing research
- Topics should be feasible and can be completed within reasonable time and resources
- Literature support: The related literature for each topic must come from actual search results and must not be fabricated
- Trend analysis: Analyze research trends based on the publication time distribution of the literature
- Differentiation: Different topics should have obvious differences and cover different research perspectives
Example workflow
用户: 我是计算机专业硕士研究生，想研究大语言模型在教育领域的应用，帮我分析选题
步骤1: 检索文献
- ArXiv: "large language models education" (50条)
- Google Scholar: "LLM education application" (20条)
- 万方: "大语言模型 教育" (50条)
步骤2: 分析文献趋势和研究空白
步骤3: 输出3-5个差异化选题建议，每个包含完整的五要素分析