Visualizing 1M+ points

This example visualizes over one million data points using dotAtlas. It shows techniques for streaming large datasets and rendering them incrementally to maintain smooth performance.

The visualization features:

  • efficient binary data format for the 2D embeddings,
  • streaming data loading with progress indication,
  • incrementally updating the visualization as the data loads,
  • improving clarity of visualization with dynamic marker size and opacity scaling.

The data in this example corresponds to over 1M abstracts of arXiv papers published between 2019 and 2024. Each point represents one abstract, thematically related abstracts are close to each other on the map. Colors correspond to the thematic cluster the abstract belongs to. The input 2D embedding and clustering were generated using the Lingo4G text analytics engine.

All the code generating this visualization is inlined into the HTML page, view page source to see the code.