Interdisciplinary Index Analyzer

A web application that measures the cross-domain impact of academic research by analyzing citation patterns and semantic similarity.

What It Does

This tool calculates a Composite Interdisciplinary Score for researchers using four equally-weighted metrics (25% each):

1. External Diversity (25%)

What it measures: How different are the papers that cite your work from your own research?

How it works:

• Fetches your top 10 most-cited papers
• For each paper, retrieves the 50 most recent citing papers
• Converts abstracts into numerical vectors using ML embeddings
• Calculates cosine distance between your papers and their citations
• Higher distance = higher diversity score

Example: If a biology paper gets cited by economics and physics papers, those citations will have high semantic distance from the original, resulting in a high diversity score.

2. Internal Diversity (25%)

What it measures: How spread out are your research topics?

How it works:

• Takes all 10 paper embeddings and computes pairwise cosine distance
• Score = average pairwise distance × 100
• Higher distance = papers cover more different topics

Example: A researcher working on both "machine learning" and "climate policy" would have high distance between papers, indicating diverse research areas.

3. Reference Diversity (25%)

What it measures: How many different fields do you draw knowledge from?

How it works:

• For each of your 10 papers, analyzes up to 50 references
• Extracts the academic field of each referenced paper
• Calculates Shannon entropy across field distribution
• More fields with balanced representation = higher score

Example: If your references span Medicine (40%), Mathematics (30%), and Computer Science (30%), you'd score higher than someone citing only Medicine (100%).

4. Bridge Score (25%)

What it measures: Are you connecting fields that don't usually talk to each other?

How it works:

• Compares fields you cite (50 refs per paper) vs. fields that cite you (50 most recent per paper)
• Identifies "bridged" fields: those citing your work that you don't cite back
• Bridge score = bridged fields / total unique fields × 100

Example: If you publish in Biology but get cited by Sociology researchers (whom you don't cite), you're bridging knowledge between disconnected fields.

Analysis Scope

Parameter	Value
Papers analyzed	Top 10 most-cited
Citing papers per paper	50 (most recent)
References per paper	50

Interpretation

Score Range	Category	Meaning
0-20%	Low	Focused research within specific domain
21-50%	Moderate	Moderate cross-disciplinary engagement
51-80%	High	Significant interdisciplinary impact
81-100%	Very High	Highly interdisciplinary work

Features

• Author Search: Search for any researcher in OpenAlex database
• Disambiguation: Select from multiple matching authors
• Top Papers Table: Ranked by External Diversity score with scrollable view
• Visualizations (8 interactive charts):
  • Metrics Overview (bullet chart)
  • Paper Similarity Heatmap (pairwise cosine similarity matrix)
  • External Diversity by Year (scatter plot)
  • Similarity Distribution (KDE)
  • Fields Referenced (bar chart)
  • Knowledge Flow (bridge analysis)
  • Citing Fields Breakdown (bar chart)
  • Top Keywords from citing papers
• Interactive HTML Export: Download full report with all charts preserved
• Caching: Session-based caching for faster repeated queries

Performance Optimizations

• Batched API Calls: Uses OpenAlex's pipe-separated ID filter to fetch 50 references in 1 call instead of 50 individual calls (96% reduction in API requests)
• Parallel Fetching: Reference Diversity and Bridge Score fetch all papers simultaneously via asyncio.gather (5-10x faster)
• Rate Limit Detection: Gracefully handles OpenAlex's 100,000 daily request limit with user-friendly error message
• Progress Logging: Step-by-step console output shows analysis progress

Tech Stack

• Frontend: Gradio
• Data Source: OpenAlex API
• ML Model: Sentence Transformers (minishlab/potion-base-32M)
• Similarity: scipy (cosine distance), scikit-learn (cosine similarity for heatmap)
• Visualization: Plotly
• Keyword Extraction: KeyBERT

Local Development

# Clone the repo
git clone https://github.com/alkat19/interdisciplinary-index-analyzer.git
cd interdisciplinary-index-analyzer

# Install dependencies
pip install -r requirements.txt

# Run the app
python interdisciplinary_app.py

The app will launch at http://localhost:7860

Acknowledgments

• OpenAlex for the open academic data API
• Sentence Transformers for efficient embeddings
• KeyBERT for keyword extraction