---
title: "CiteGraph | Scholarly Network Intelligence"
emoji: "🔬"
colorFrom: green
colorTo: blue
sdk: static
app_file: index.html
pinned: false
license: mit
short_description: "Citation network analysis and literature mapping"
tags:
  - research
  - bibliometrics
  - network-analysis
  - academic-research
  - data-visualization
  - javascript
---

# CiteGraph | Scholarly Network Intelligence

CiteGraph is a **fully offline bibliometric analysis and citation-network explorer** developed as a single-page static application. It visualizes links between scholarly publications and computes transparent, reproducible metrics that describe a chosen literature corpus.

**Author:** Rodrigo Pantoja | **Release:** v1.0 | **Project:** #43

## Capabilities

- **Overview:** publication count, citation links, citation coverage, topic distribution and an interactive citation network.
- **Network Explorer:** click nodes to inspect a paper's references and citations received *within the uploaded corpus*. Search by paper title, author or ID, and filter by topic.
- **Influence Rankings:** compute incoming citations and directed PageRank (damping 0.85); inspect top publications and scores.
- **Themes & Timeline:** examine publication trends, bibliographic coupling, co-citation links and topic metadata.
- **Audit & Methods:** review source-data validation, graph structure, integrity indicators and methodological cautions.
- **Data & Exports:** import papers and citations as UTF-8 CSV files; download summary Markdown, JSON audit report, enriched papers and raw citation links.

### Included example collections

The application includes three **synthetic and entirely fictional datasets**:

1. Responsible AI Research
2. Climate Resilience Systems
3. Cross-Border Education

These examples are *not actual published papers*, and their authors, citation links and scores must never be quoted as real scholarship or empirical evidence.

## How to use

1. Open the **App** tab of this Space.
2. Choose one of the three sample collections from the selector at the top.
3. Use **Network Explorer** to select research-paper nodes.
4. See **Influence Rankings**, **Themes & Timeline**, or **Audit & Methods** for other views.
5. To import your data, open **Data & Exports**, select a papers CSV and citations CSV, then press **Import & validate**.
6. Download audit reports and result tables from the same tab.

## CSV schema

`papers.csv` must contain the following columns:

```csv
id,title,year,topic,authors
A1,A research paper,2021,Methods,A. Author
A2,Further investigation,2022,Methods,B. Author
A3,New evidence,2023,Practice,C. Author
```

`citations.csv` must contain:

```csv
source,target
A2,A1
A3,A1
A3,A2
```

**Direction matters:** source cites target.

Datasets must contain 3 to 120 publications, up to 600 links, unique publication IDs, valid publication years (1950–2035) and no dangling or duplicate links. Self-references and references to papers published in a future year are rejected. CSV uploads are limited to 1.5 MB per file. A rejected import **does not replace** the current dataset.

## Methodology

- **In-degree:** the number of citation links pointing to a paper from *inside this corpus*.
- **Out-degree:** the number of reference links going from a paper to other papers *inside this corpus*.
- **PageRank:** a normalized iterative centrality score based on directed citation links, damping factor 0.85, with uniform redistribution from dangling nodes.
- **Internal h-index:** the h-index calculated from **internal citation counts only**, not Google Scholar, Scopus or other databases.
- **Citation coverage:** share of corpus publications with at least one incoming internal citation.
- **Directed density:** observed citation links divided by `N × (N − 1)`.
- **Bibliographic coupling:** shared references between two papers (a Jaccard score is supplied).
- **Co-citation:** papers cited by the same source publications.
- **Weak connectivity:** size of the largest undirected component as a fraction of publications.
- **Topics:** labels from the supplied CSV, not automatically learned scholarly research communities.

## Scientific limitations

CiteGraph is a reproducible descriptive research tool, **not a global citation index**. Its rankings are conditional on the completeness and representativeness of the submitted papers and references. Citation counts and centrality scores are not measures of scientific quality, research impact beyond the corpus or causal influence. In particular, do not interpret synthetic demo graphs as empirical findings.

There are no backend requests, paid APIs, tracking libraries, model inference endpoints or CDN dependencies. Processing takes place entirely in the browser. Reloading the page resets imported data to a demo dataset.

## Technical setup

- Hugging Face SDK: `static`
- Entry point: `index.html`
- Languages: HTML, CSS, JavaScript
- Dependencies: none
- Publishing files: `README.md`, `index.html`, `LICENSE`
- License: MIT

## Tests

The project source distribution includes automated tests for CSV parsing, graph validation, PageRank, centrality, data import, all six screens, download formats and mobile interaction. Testing is performed against demo and small known-answer datasets. A deployed Space should also be checked manually after uploading.