By the way, people use Markov chains to generate meaningless texts. Modern smartphones use word pair statistics from others’ and your own sentences to suggest the next word. Try writing anything offensive – it will continue correctly;) so, a Markov chain is a sequence of such assumptions based on the statistics of combinations. In my almost childhood, they tried to deceive search engines by generating placeholder pages with supposedly real text and advertising links inside (they’ve long caught on to this trick). Also, I used this mechanism to create random words, which can be read effortlessly (filled the dictionary with them and obfuscated data from real logs for blog publication).
There’s another interesting application. You probably didn’t know that Markov chains underlie Google’s method of sorting search results. If we draw parallels with words, then here the pairs are often-used words – pairs of sites linked by hyperlinks. There is an interesting property of the Markov chain: if it is long enough, and there is enough data, then in the “long distance,” the probability of the outcome (words in case of suggestions) does not depend on the word you started writing with, assuming we’re talking about a long chain. So, the higher the probability, the higher the pagerank, the closer to the start the result will be. Well, it’s understood that Google has seriously modified the algorithm, but the principles of this stuff with links and Google’s Pagerank are very similar;)
https://meduza.io/shapito/2018/07/13/prochital-na-meduze-chto
Hi, I'm Rauf Aliev. 