Jev can only answer multiple-choice questions. Ask it something and watch it answer anyway, one word at a time.
Each word takes a few questions, so answers come out slowly.
The questions for each word will appear here.
How the questions work
Jev, from TypeSafe AI, is a decision model. It can't write text. You give it some context and a multiple-choice question, and it returns a probability for every option.
So every English word is placed at the end of a path of named choices. The first choice is grammar: noun, verb, adjective, pronoun, punctuation, and so on (the universal part-of-speech tags). For nouns, verbs and adjectives, the next choices follow WordNet, Princeton's dictionary of meanings: what the word is about, then what kind of thing it is, narrowing down until one word is left. Last, it picks the form: walk, walks, walked or walking.
Each question sends Jev your question, the answer so far, and the choices already made for this word. The percentages are Jev's, and the page follows the most likely option.
Names and words that aren't in the tree get spelled out one letter at a time. When the answer is complete, Jev picks "end of answer".