Star Wars Edges — Dataset Explorer

A React + D3 single-page visualization that loads the starwars_edges.csv project dataset at runtime and renders:

1. A dataset summary (name, file size, row count, column count).
2. A preview table of the first 10 rows with all column headers.
3. This README file, documenting the dataset schema.

Loading

The dataset is loaded at runtime from its project-dataset virtual path: /data/starwars_edges.csv

Schema

The dataset is a directed, weighted edge list representing character interactions across all Star Wars films (Episodes I–VI). Each row denotes that character source interacts with character target in a given episode, with value representing the intensity (e.g., number of scenes, lines spoken together, or other normalized proximity metric).

All edge values are non-negative integers. The dataset is sparse: most character pairs appear only once or twice across the full corpus.

Visualization Ideas

1. Episode-Binned Force-Directed Network

   • Question: How do character communities evolve across episodes?
   • Fields: episode, source, target, value
   • Encoding: Nodes = characters (radius ∝ degree centrality), edges = weighted (opacity/width ∝ value), color = episode (1–6), layout via force-directed.
   • Interaction: Toggle episodes on/off; hover to show edge weight and characters; click to highlight ego-network of a selected node.
   • Note: Render 6 separate force layouts (one per episode) or animate transitions between episodes.

2. Aggregated Adjacency Matrix (Heatmap)

   • Question: Which character pairs interact most often, and in which episodes?
   • Fields: episode, source, target, value
   • Encoding: Rows/columns = sorted characters (by total degree), cell = sum of values across episodes, color = intensity (log-scaled), shape = binary presence.
   • Interaction: Hover to show (source, target, total weight, episode breakdown); click a cell to filter table to those edges.
   • Note: Use hierarchical clustering (e.g., hierarchical edge bundling on matrix) to reorder rows/columns.

3. Episode-Parallel Arc Diagram

   • Question: How do interactions cluster spatially within each episode?
   • Fields: episode, source, target, value
   • Encoding: Horizontal layout per episode; nodes on x-axis (alphabetical/centralized), arcs over axis (thickness ∝ value), color = episode.
   • Interaction: Hover arcs to show edge weight; click episode label to highlight only that arc layer.
   • Note: Ideal for visualizing temporal proximity (e.g., sorting by narrative order).

4. Sankey Flow Diagram

   • Question: How do characters flow between episodes in terms of interaction volume?
   • Fields: episode, source, target, value
   • Encoding: Nodes = characters per episode (left-right across episodes), flows = weighted edges, color = source character group.
   • Interaction: Hover flows for value; toggle episodes to isolate segment.
   • Note: Require node expansion to avoid excessive edge bundling; consider splitting nodes across episodes.

5. Chord Diagram (Episode Aggregated)

   • Question: What is the pattern of cross-character interaction strength within and across episodes?
   • Fields: episode, source, target, value
   • Encoding: Circular layout; nodes = top N characters; ribbons = aggregated value between pairs, fill = symmetric weight.
   • Interaction: Hover to show source, target, aggregated weight, episode distribution.
   • Note: Aggregate across episodes first; use symmetric matrix (sum of A→B + B→A) if direction is not essential.

6. Small Multiples Timeline (by Episode)

   • Question: How does the character interaction network evolve per episode?
   • Fields: episode, source, target, value
   • Encoding: 6 small force-directed or edge-bundled networks, one per episode; node size ∝ episode-degree; edge width ∝ episode-weight.
   • Interaction: Synced brushing across plots; click episode to highlight in table.
   • Note: Align node positions (e.g., spring embedder with fixed seed) for comparability.

7. Character Rank-Order Bar Chart

   • Question: Who are the most central characters in each episode?
   • Fields: episode, source, target, value
   • Encoding: Bar chart with y-axis = character, x-axis = sum(incoming value), bars grouped/stacked by episode.
   • Interaction: Sort by total, episode, or alphabetically; hover to show breakdown by episode.
   • Note: Use horizontal bars for long character names; invert axis if > 30 characters.

8. Weight Distribution Histogram

   • Question: What is the distribution of interaction strengths across the dataset?
   • Fields: value
   • Encoding: Histogram bins for value counts; overlay density curve if needed.
   • Interaction: Click bin to filter table by edge weight range.
   • Note: Use log-scale x-axis due to heavy-tailed distribution.

9. Episode Heatmap Matrix (Character × Episode)

   • Question: Which characters appear in which episodes, and how strongly?
   • Fields: episode, source, target, value
   • Encoding: Rows = top N characters, columns = episodes, cell = max(incoming+outgoing value) for that character/episode, color intensity.
   • Interaction: Hover to show character’s total episode contribution; click to highlight all related edges.
   • Note: Normalize per character/episode if scale differences dominate.

10. Cumulative Weight Timeline

    • Question: How do cumulative interaction volumes build across episodes?
    • Fields: episode, source, target, value
    • Encoding: Line/area chart: x = episode, y = cumulative sum of edge weights, lines = top characters.
    • Interaction: Toggle characters; hover for exact cumulative value.
    • Note: Could show per-character or aggregated over all characters.

11. Edge Duration Stacked Bar (Episode-Specific)

    • Question: Which character pairs co-appear longest in each episode?
    • Fields: episode, source, target, value
    • Encoding: Horizontal stacked bars: one row per (source, target) pair, color = episode, segment length = value.
    • Interaction: Click pair to highlight all episodes for that edge.
    • Note: Limit rows to top 30 pairs to avoid clutter.

12. Character Trajectory Graph

    • Question: How does a character’s influence (degree, weighted in/out) change across episodes?
    • Fields: episode, source, target, value
    • Encoding: Line chart: x = episode, y = degree/weight metric, separate line per character.
    • Interaction: Hover for values; select character to highlight in network views.
    • Note: Show both in-degree and out-degree as separate series.

13. Interactive Edge List Table with Filters

    • Question: How can users explore raw edges by custom criteria?
    • Fields: episode, source, target, value
    • Encoding: Tabular view with column filters (episode dropdown, character search, value slider).
    • Interaction: Filter in real time; export selected rows; link to visualizations (e.g., click row to show edge in network).
    • Note: Use Debounce on filters; support regex on character names.

14. Episode Co-occurrence Network (Character × Episode)

    • Question: Which episodes share the same core character groups?
    • Fields: episode, source, target, value
    • Encoding: Bipartite network: one set = characters, other = episodes, edges = character appears (value > 0), weight by sum of incident edge values.
    • Interaction: Click episode to show character set; click character to show episode list.
    • Note: Project to character-only (if same episode co-occurrence desired) via matrix multiplication.

15. Radial Edge Bundling (Episode → Characters)

    • Question: Do interaction patterns cluster around central figures per episode?
    • Fields: episode, source, target, value
    • Encoding: Radial layout: center = episode hub, radial spokes = characters (sorted by episode centrality), edges = weighted arcs to other characters.
    • Interaction: Expand character by clicking; show tooltip with centrality metrics.
    • Note: Simplify by using only top 20% degree characters per episode.

16. Topological Sorting for Narrative Chronology

    • Question: Can we infer character appearance sequence or centrality order across episodes?
    • Fields: episode, source, target, value
    • Encoding: Time-series edge diagram (x = episode, y = character name), edges = source→target over episodes.
    • Interaction: Hover edges to show weight; group characters by episode debut or cumulative weight.
    • Note: Order characters by debut episode or total episode participation.

17. Betweenness Centrality Animated Map

    • Question: Which characters act as bridges across episodes?
    • Fields: episode, source, target, value
    • Encoding: Bar chart of betweenness centrality (computed per episode), bars animated over time; color = betweenness percentile.
    • Interaction: Toggle episode; hover to show character name and value.
    • Note: Precompute centrality per episode to avoid runtime overhead.

18. Episode Pair Sankey Flow

    • Question: How do character sets overlap between consecutive episodes?
    • Encoding: Nodes = characters in episode i, edges to characters in episode i+1, flow = shared edge weight.
    • Fields: episode, source, target, value
    • Interaction: Hover to show overlap strength; click pair to isolate episode jump.
    • Note: Focus on sequential episode pairs (e.g., 1→2, 2→3, etc.).

19. Weighted Edge Cumulative Curve

    • Question: What proportion of total interaction weight comes from top edges?
    • Fields: value
    • Encoding: Pareto-style curve: x = edges sorted by weight (desc), y = cumulative % of total weight.
    • Interaction: Hover to show weight threshold and % coverage; click point to filter table.
    • Note: Add reference lines (e.g., 80% of weight from top 10% edges).

20. Character Interaction Timeline Heatmap

    • Question: When do character pairs first/last appear together?
    • Fields: episode, source, target, value
    • Encoding: Rows = top edge pairs, columns = episodes 1–6, cell = max weight in episode or binary first/last.
    • Interaction: Click pair to highlight in network or show all edges for that pair.
    • Note: Use binary for first appearance; color by value for repeated interactions.