How to make a portrait according to words?

Question

Algorithms

I have one person's speech. I need to make a more unusual report from his words. So far, I have come up with a word cloud based on the frequency of repeated words, and a map of photographs (often repeating word is replaced by a photo with a size proportional to the repetition of the word). What is a word cloud, I think many people know.
As a result, I would like to calculate what kind of image he has, which hero from Star Wars or the Lord of the Rings his speech is similar to. But how to do that?
Parse the dialogues from the Star Wars book, group them by repeated words and find the array of words that most closely matches the person .... It’s somehow difficult and it will be navryatli for sure.
Does anyone have an idea how this can be done? At least in which direction to look?

Answer 1

It seems to me that a person's speech, unless it is the speech of a narrow specialist at the workplace or at a lecture, is in fact almost always about the same things. The specificity of a particular person lies in the choice of synonyms, introductory words, the most frequent schemes for constructing sentences. Turning these differences into pictures is impossible. They (differences) will be erased: how do you draw "cool", "cool" and "cool" differently?
If, however, it is the words that are left in the report, and not their images, then this is a fairly well-known task of comparing the text under study with the corpora of the most commonly used words. The means of mathematical statistics are deviations observed in the speech of this person, relative to the common vocabulary, and some kind of report or visualization is built from them:
habrahabr.