| name | academic-search |
| description | Techniques for querying academic databases (ArXiv, Semantic Scholar, PubMed) and code repositories (GitHub) |
Academic Search Skill
Overview
This skill provides strategies and tool usage patterns for effectively searching academic literature. It focuses on precision searching to find high-quality, relevant papers.
Search Strategies
1. Keyword Expansion
Start with a broad topic and generate synonyms.
- Topic: "Large Language Models in Healthcare"
- Keywords: "LLM", "Generative AI", "NLP", "Clinical Decision Support", "Electronic Health Records"
2. Boolean Operators
Use standard operators to refine searches:
AND: narrow results (e.g., LLM AND "clinical trial")
OR: broaden results (e.g., cancer OR oncology)
NOT: exclude specific terms (e.g., virus NOT "computer virus")
3. Database Specifics
ArXiv (Computer Science, Physics)
- Best for: Preprints, cutting-edge AI research.
- Search Tip: Use category filters (e.g.,
cat:cs.AI, cat:cs.CL).
- Sorting: Sort by "Submitted Date" to see the absolute latest work.
Semantic Scholar (General Science)
- Best for: Finding connected papers, traversing citation graphs.
- Search Tip: Use "Highly Influential Citations" to find seminal papers.
- Filtering: Filter by "Has PDF" to ensure full-text access.
GitHub (Open Source Code)
- Best for: Finding implementation code, frameworks, tools.
- Search Tip: Include "implementation", "code", or specific languages (e.g.,
LLM language:python).
- Sorting: Sort by "stars" to find the most popular repositories.
Usage
Use the provided Python scripts to fetch data.
1. ArXiv Fetcher
python .research-agent/skills/academic-search/scripts/arxiv_fetcher.py --query "Machine Learning" --max_results 5
2. Semantic Scholar Fetcher
python .research-agent/skills/academic-search/scripts/scholar_fetcher.py --query "Transformer Architecture" --limit 5
3. GitHub Fetcher
python .research-agent/skills/academic-search/scripts/github_fetcher.py --query "Research Agent" --limit 5 --sort stars
Verification
- Citation Count: High citations usually indicate impact.
- Venue: Check the conference (NeurIPS, ICML) or journal impact factor.