Abstract Summarizer

from scripts.summarizer import AbstractSummarizer

summarizer = AbstractSummarizer()

# Generate from PDF
abstract = summarizer.summarize(
    source="paper.pdf",
    format="structured",  # structured, plain, or executive
    word_limit=250,
    discipline="biomedical"  # affects terminology handling
)

print(abstract.text)
# Output: Background → Objective → Methods → Results → Conclusion

**Background**: [Context and problem statement]
**Objective**: [Research goal and hypotheses]
**Methods**: [Study design, sample, key methods]
**Results**: [Primary findings with statistics]
**Conclusion**: [Implications and significance]

---
Word count: 247/250

# Extract and verify quantitative results
quant_results = summarizer.extract_quantitative(
    text=paper_content,
    priority="high"  # keep all numbers vs. representative samples
)

# Validate against original
validation = summarizer.verify_accuracy(
    abstract=abstract,
    source=paper_content
)

# Biomedical paper
python scripts/main.py --input paper.pdf --field biomedical

# Physics paper  
python scripts/main.py --input paper.pdf --field physics

# Social science paper
python scripts/main.py --input paper.pdf --field social-science

from scripts.batch import BatchProcessor

batch = BatchProcessor()

# Process directory of papers
summaries = batch.summarize_directory(
    directory="literature_review/",
    output_format="csv",  # or json, markdown
    include_metadata=True  # title, authors, year
)

# Generate review matrix
matrix = batch.create_summary_matrix(summaries)
matrix.save("review_matrix.csv")

{
  "paper_type": "clinical_trial",
  "key_elements": [
    "Study design (RCT, cohort, case-control)",
    "Population (n, inclusion/exclusion)",
    "Intervention details",
    "Primary endpoint",
    "Key results (efficacy, safety)",
    "Clinical significance"
  ],
  "emphasis": "P-values, confidence intervals, adverse events"
}

**Background**: Current treatments for X disease have limited efficacy.
**Objective**: Evaluate Drug Y's safety and efficacy in patients with X.
**Methods**: Double-blind RCT (n=342) comparing Drug Y vs placebo for 12 weeks.
**Results**: Primary endpoint achieved (67% vs 32% response, p<0.001, OR=4.2). 
            Adverse events mild (headache 12%, nausea 8%).
**Conclusion**: Drug Y significantly improves outcomes with acceptable safety profile.

{
  "paper_type": "basic_science",
  "key_elements": [
    "Research question/hypothesis",
    "Model system (cell line, animal, in vitro)",
    "Key methods (CRISPR, Western blot, etc.)",
    "Mechanistic findings",
    "Biological significance"
  ],
  "emphasis": "Molecular mechanisms, pathway diagrams"
}

**Background**: The role of Protein X in Disease Y progression is unknown.
**Objective**: Determine if Protein X regulates Pathway Z in Disease Y.
**Methods**: CRISPR knockout in cell lines, Western blot analysis, mouse model.
**Results**: Protein X deletion reduced Pathway Z activation by 78% (p<0.01). 
            In vivo, knockout mice showed 45% less disease progression.
**Conclusion**: Protein X is a critical regulator of Pathway Z and potential therapeutic target.

{
  "paper_type": "meta_analysis",
  "key_elements": [
    "Search strategy and databases",
    "Number of studies included",
    "Total sample size",
    "Pooled effect size",
    "Heterogeneity assessment",
    "Quality of evidence"
  ],
  "emphasis": "I² values, funnel plots, GRADE assessment"
}

**Background**: Previous trials of Intervention X show conflicting results.
**Objective**: Systematically evaluate efficacy through meta-analysis.
**Methods**: PRISMA-guided search of PubMed, Embase, Cochrane (through 2024). 
            23 RCTs (n=4,847) met inclusion criteria.
**Results**: Significant benefit observed (SMD=0.42, 95% CI [0.28, 0.56], p<0.001). 
            Moderate heterogeneity (I²=45%). Quality: moderate.
**Conclusion**: Intervention X shows modest efficacy with moderate certainty evidence.

{
  "paper_type": "methodology",
  "key_elements": [
    "Problem with existing methods",
    "Novel approach description",
    "Key innovations",
    "Performance benchmarks",
    "Comparison to state-of-the-art"
  ],
  "emphasis": "Accuracy, speed, scalability metrics"
}

**Background**: Current algorithms for Problem X are computationally expensive.
**Objective**: Develop efficient method with improved accuracy.
**Methods**: Novel graph neural network architecture with attention mechanism. 
            Validated on 5 benchmark datasets.
**Results**: 3.2× faster than current methods with 12% accuracy improvement 
            (p<0.001). Scales to datasets with 10M+ nodes.
**Conclusion**: Method achieves superior performance with practical computational requirements.

# Step 1: Extract text from PDF
python scripts/extract.py --input paper.pdf --output paper.txt

# Step 2: Generate structured abstract
python scripts/main.py \
  --input paper.txt \
  --field biomedical \
  --format structured \
  --word-limit 250 \
  --output abstract.md

# Step 3: Verify accuracy
python scripts/verify.py \
  --abstract abstract.md \
  --source paper.txt \
  --check-quantitative \
  --output verification_report.txt

# Step 4: Adapt for specific journal
python scripts/adapt.py \
  --abstract abstract.md \
  --journal "nature_medicine" \
  --output submission_abstract.txt

from scripts.summarizer import AbstractSummarizer
from scripts.validator import AccuracyValidator

# Initialize
summarizer = AbstractSummarizer()
validator = AccuracyValidator()

# Summarize
with open("paper.pdf", "rb") as f:
    abstract = summarizer.summarize(
        source=f,
        discipline="clinical",
        word_limit=250
    )

# Verify numbers are accurate
is_accurate = validator.check_quantitative(
    abstract=abstract,
    source_pdf="paper.pdf"
)

if is_accurate:
    abstract.save("final_abstract.txt")
else:
    discrepancies = validator.get_discrepancies()
    print(f"Review needed: {discrepancies}")