In April, I wrote a text generator in Python – find here the article in French – long live Markov Chains! I can now generate sentences by feeding my generator with my texts – here in English.
Roughly, my algorithm will compute the probability of my writing some words in my texts, after Word 1 and Word 2. For instance, in my texts, after “It is”, let’s say there is a 10% chance of having “obvious”, 20% “raining”, 25% “a”, and so on. Thanks to this data my generator will choose words according to those probabilities. If my generator first chooses “It is” as the beginning of a sentence, there is a 10% chance the next word will be “obvious”, 20% “raining”, etc.
The resulting sentences are often funny, always very surreal.
Selected generated sentences:
“I will be me. I was twelve, the trend in art was realistic descriptions of the deep maze inside… Recovered memories suddenly propagate, What he recalls from her throws him in a modernist museum.”
Continue reading Texts I could have written – Generated Sentences (Markov Chains)