Academic Research Use Case
Use Zorora’s deep research capabilities for comprehensive academic research and literature reviews.
Scenario
You’re writing a research paper on “Recent advances in transformer architectures for natural language processing” and need to:
- Find recent papers on the topic
- Explore citation trails
- Cross-reference claims
- Synthesize findings with citations
(Placeholder - Add screenshot showing academic research results with citations and credibility scores)
Step-by-Step Guide
Step 1: Start Research Query
Terminal:
zorora
[1] ⚙ > What are the latest developments in transformer architectures for natural language processing?
Web UI:
- Open
http://localhost:5000 - Enter query: “latest developments in transformer architectures for natural language processing”
- Select depth: Thorough (for comprehensive research)
- Click “Start Research”
API:
from engine.research_engine import ResearchEngine
engine = ResearchEngine()
state = engine.deep_research(
"latest developments in transformer architectures for natural language processing",
depth=3 # Thorough depth for academic research
)
Step 2: Review Sources
Academic Sources Found:
- arXiv papers on transformer architectures
- PubMed articles (if biomedical relevance)
- Google Scholar results
- CORE repository papers
Source Credibility:
- High credibility (0.7-1.0): Peer-reviewed journals, well-cited papers
- Medium credibility (0.4-0.7): Preprints, recent papers
- Low credibility (0.0-0.4): Unverified sources
Review Sources:
for source in state.sources_checked:
if source.source_type == "academic":
print(f"{source.title}: {source.credibility_score:.2f}")
print(f" URL: {source.url}")
Step 3: Explore Citation Trails
Citation Following:
- Zorora automatically follows citations from initial sources
- Explores cited papers (up to 3 hops deep)
- Builds citation graph
- Identifies key papers
Access Citation Graph:
# Citation relationships are stored in research state
# Access via research data
research_data = engine.load_research(state.started_at.strftime("%Y%m%d_%H%M%S"))
citation_graph = research_data.get('citation_graph', {})
Step 4: Cross-Reference Claims
Grouped Claims:
- Zorora groups similar claims across sources
- Counts agreement
- Identifies consensus
- Highlights disagreements
Review Findings:
for finding in state.findings:
print(f"Claim: {finding.claim}")
print(f"Agreement: {finding.agreement_count} sources")
print(f"Sources: {finding.sources}")
Step 5: Synthesize Findings
Synthesis Includes:
- Comprehensive answer with citations
- Key findings highlighted
- Areas of consensus
- Areas of disagreement
- Confidence levels
Access Synthesis:
print(state.synthesis)
Step 6: Save Research
Save for Later:
research_id = engine.save_research(state)
print(f"Research saved: {research_id}")
Access Later:
research_data = engine.load_research(research_id)
Example Output
Synthesis
Based on recent research, transformer architectures for natural language processing have seen significant developments in several areas:
**Efficiency Improvements:**
- Sparse attention mechanisms [Academic: arXiv:2024.12345]
- Model compression techniques [Academic: arXiv:2024.67890]
- Efficient training methods [Web: example.com/article]
**Architecture Innovations:**
- Multi-head attention variants [Academic: arXiv:2024.11111]
- Position encoding improvements [Academic: arXiv:2024.22222]
**Key Consensus:**
- Sparse attention is widely accepted as effective [8 sources agree]
- Model compression maintains quality [6 sources agree]
**Areas of Disagreement:**
- Best position encoding method [3 sources disagree]
Sources
High Credibility Sources:
- “Efficient Transformers: A Survey” (arXiv:2024.12345) - 0.85
- “Sparse Attention Mechanisms” (Nature ML) - 0.90
- “Transformer Compression Techniques” (ICML 2024) - 0.88
Medium Credibility Sources:
- “Recent Transformer Advances” (arXiv:2024.67890) - 0.65
- “Transformer Architecture Survey” (arXiv:2024.11111) - 0.70
Best Practices
Query Formulation
Be Specific:
- “Recent advances in transformer architectures” ✓
- “AI stuff” ✗
Include Context:
- “Transformer architectures for natural language processing” ✓
- “Transformers” ✗
Use Research Keywords:
- What, Why, How, Tell me
- Automatically triggers research workflow
Depth Selection
Quick (depth=1):
- Initial research
- Fast answers
- Overview of topic
Balanced (depth=2):
- Thorough research
- Citation exploration
- Academic focus
Thorough (depth=3):
- Comprehensive research
- Deep citation trails
- Maximum coverage
Source Evaluation
Check Credibility:
- Prioritize high-credibility sources
- Verify claims from multiple sources
- Note areas of disagreement
Follow Citations:
- Use source URLs to verify
- Explore citation graph
- Check original sources
Consider Context:
- Note publication dates
- Consider source bias
- Verify claims independently
Advanced Usage
Multiple Research Queries
Related Topics:
[1] ⚙ > What are the latest developments in transformer architectures?
[2] ⚙ > How do attention mechanisms work in transformers?
[3] ⚙ > What are efficient training methods for transformers?
Compare Results:
- Load multiple research results
- Compare findings
- Identify trends
Citation Graph Analysis
Explore Relationships:
# Load research with citation graph
research_data = engine.load_research(research_id)
citation_graph = research_data.get('citation_graph', {})
# Identify key papers (most cited)
key_papers = identify_key_papers(citation_graph)
Cross-Reference Verification
Verify Claims:
# Find claims with high agreement
high_agreement = [f for f in state.findings if f.agreement_count >= 5]
# Find claims with disagreement
disagreement = [f for f in state.findings if f.agreement_count <= 2]
Troubleshooting
No Academic Sources Found
Problem: Research returns no academic sources
Solution:
- Check query phrasing
- Try different keywords
- Verify academic search is working
- Check API connectivity
Low Credibility Scores
Problem: All sources have low credibility
Solution:
- This may be normal for niche topics
- Check source URLs manually
- Try different query phrasing
- Consider using Balanced/Thorough depth
Citation Graph Empty
Problem: Citation graph not populated
Solution:
- Use depth > 1 (Balanced or Thorough)
- Verify citation following is enabled
- Check research depth setting
See Also
- Research Workflow - Deep research guide
- Technical Concepts - Pipeline architecture
- API Reference - Programmatic access