Word cloud generator

Two decisions do all the work. The first is the stop list: without it, every word cloud in the world reads THE, AND, OF in the largest type, because those are the most common words in English by an enormous margin and they carry no information about the text. Filtering them is what makes the picture about the subject rather than about English.

The second is the size curve. Word frequencies follow a power law — in most texts the commonest word appears about twice as often as the second and ten times as often as the tenth — so scaling font size linearly with count makes the top word enormous and everything below it unreadable. Sizes here follow the square root of the frequency, which compresses that range into something the eye can actually compare.

The layout is an Archimedean spiral: each word starts at the centre and walks outward until it finds a space nothing else occupies. That is the standard approach and it is why clouds look organic rather than gridded. Words that never find a space are reported rather than silently dropped, because a cloud that has quietly discarded a third of your text is misleading.

Worth saying plainly: a word cloud is a decorative summary, not an analysis. It shows frequency and nothing else — not sentiment, not importance, not context. A word appearing often because it is being criticised looks identical to one appearing often because it is being praised.