Physics Concept Extractor

   Is there an automated or semi-automated way to extract the Core Concepts of a given Discipline? Using Physics as an example, here we present the results of our preliminary investigations.

   The aim of this project is to, first, identify some analog of “concepts” throughout loosely structured lectures and course material, and second, determine the relationships between these concepts. Feel free to experiment on your own using this visualization!

This data was generated by a neural network trained to determine the relationships between words, and to assign those words positions in a 300 dimensional space in such a way that strongly related words are placed in close proximity to one another. The interactive 3D visualization on this page approximates the relative distances of words in that 300 dimensional space such that words close to one another in the 300 dimensional space are roughly close to one another in this 3D visualization.


  • You can hover over a point on the graph to see the word it represents.
  • The Color of a point represent the topics (legend displayed in left hand panel) to which its corresponding word belongs. The more vibrant the color, the more that word belongs exclusively to that topic.
  • Clicking on a point will select it, and desaturate points corresponding to unrelated words.
  • Selecting a word will also show a list of its most related words on the right hand panel.
    • The right hand panel also includes a search filter, so you can find words of interest.
  • Selecting a word will update the colored-boxes on the left hand panel to indicate how strongly the selected word is associated with the corresponding topic.
    • The left hand panel includes a slider which can increase or decrease the number of topics into which the words are categorized.
    • The bottom of the left hand panel can be used to select different mappings of the 300 dimensional space into the 3dimensional visualization.
  • Controls: Left-click and drag to rotate the graph. Right-click and drag to pan. Scroll to zoom.

(To read more about the rationale, methodology, and technology behind this experiment, please download this document.)

Instructions & Info >>

Number of topics: 4
topic 1
topic 2
topic 3
topic 4
topic 5
topic 6
topic 7
topic 8
topic 9
Projection Distance Bias