Build and interact with a visualization of your own knowledge
Answer questions to watch our system's model of you update in real time
Click info in the upper right to learn more, or read our research paper.
Knowledge Mapper uses text embedding models to create a semantic coordinate system from 250,000 Wikipedia articles. Each article is embedded into a high-dimensional vector space and projected onto a 2D map where related concepts cluster together — the articles serve as guideposts that define the terrain of human knowledge.
As you answer questions, an adaptive estimator interpolates your knowledge across nearby concepts using radial basis functions. Each answer creates a ripple of evidence: getting a question right suggests you likely know related topics nearby, while a wrong answer reveals gaps. The system uses uncertainty-weighted selection to pick the most informative next question, efficiently mapping your understanding with as few questions as possible.
Because the map lives in embedding space, knowledge demonstrated in one area (e.g., Mathematics) provides evidence about related areas (e.g., Probability & Statistics). The result is a high-resolution heatmap of your conceptual knowledge — not a single score, but a spatial portrait revealing both your strengths and gaps.
The questions are intended to be difficult. Each question tests multiple aspects of knowledge: conceptual understanding, familiarity with relevant terminology, reasoning, and reading comprehension. Don’t worry if many feel challenging — the system learns from both correct and incorrect answers to build a complete picture of your knowledge.
Research paper: Text embedding models yield detailed conceptual knowledge maps derived from short multiple-choice quizzes
Source code: github.com/ContextLab/mapper
| Esc | Dismiss quiz panel or modal |
| 1–4 | Quick answer (options 1/2/3/4) |
| Tab | Cycle through interactive controls |
| Scroll / Pinch | Zoom the map |
| Click + Drag | Pan the map |
Click on the correct response to answer each question. Use the number keys 1–4 for quick answers.
All computation runs in your browser. No data is sent to any server. Your progress is saved locally and can be exported or reset at any time.
Note: The knowledge estimates assume that your responses reflect genuine effort. Randomly clicking through questions without thinking will generate estimates that aren’t useful or meaningful.
This work was supported by NSF award number 2145172.