conndict is short for connected dictionaries.

This project attempts to answer this question:

Starting from one word in the dictionary, you may look up any word that occurs in that word's definition (and so on, recursively). What word should you begin with to learn the most words?

This is evidently a graph traversal problem where words are vertices and some word w being in the definition of another word m implies a connection m -> w. Then we can answer this question with any graph traversal algorithm.

I got an API key from Merriam-Webster for looking up definitions and figured out how to parse their results. I then wrote a simple web page where you can start with any word and click to expand words in its definition to explore the graph manually. This can get pretty big pretty quickly.

For running it:

1. Get an API key here and put it in secrets.go
2. go build
3. ./conndict -serve
4. Navigate to localhost:8000/view/[your starting word here]

This is a lot of fun and really cool to look at, but it doesn't really answer the original question.

I want to know what the answer to the original question is. This will probably entail dynamically building my own database of definitions (since Merriam-Webster is slow). There's a lot of storage commitment, then writing the code, and then letting that code run for however long it needs.

Since each query to Merriam-Webster takes about 1 second, and a baseline estimate for words in English is 171,000, running the program alone could take anywhere between two days and two weeks.