What it is
ConceptMapper is a thinking aid for anyone with a long list of ideas. You enter your ideas with short descriptions and it lays them out on a map where similar ideas sit near each other. Its favorite trick is the "distant analogy bridge": two ideas that are about different things but work the same way, which can spark a combination you would not have seen. A second view helps you choose by plotting impact against effort.
How it works
- Add ideas one at a time or paste a batch, each with a title and a description.
- Press Analyze.
- Each idea is turned into a numeric "meaning fingerprint" (an embedding) and placed on a 2D map.
- The app finds pairs that are far apart in topic but similar in function, and draws a bridge between them.
- Switch to the portfolio view: each idea gets an impact and effort score, and the best trade-offs are highlighted.
What exists today
- A local web app with idea entry, the meaning map and the portfolio chart
- Saved ideas in a small local database
- The analogy-bridge finder and the best-trade-off line
- Experimental variants with an interactive dashboard and a 3D map
- An earlier notebook that applied the same approach to clustering customer feedback
What isn't there yet
- Never packaged; runs only on a computer with Python set up
- No hosted version and no demo data set
- No tests
- Impact and effort scoring is simple and not explained in the app
- No import from other tools, no export of the map
Under the hood
- Python with Streamlit for the interface and SQLite for storage
- A small open sentence-embedding model (MiniLM) runs locally, no AI service needed
- Maps are flattened to 2D with PCA; "same kind of thing" is measured with word-weighting (TF-IDF) on descriptions
- About 200 lines of app code plus small modules for storage, embeddings, signals and scoring