Little clojure lib for creating pieces of random text that tries to disguise as the real text. Example of use: placeholder text for a webpage in development.
Obvious alternative is lorem ipsum. I however wanted to implement something similar myself, and to do it in clojure as an exercise of sorts.
To create a random text that at least vaguely resembles real human english text I needed to make sure artificial text possesses at least some qualities characteristic of the real text.
Firstly, i found data on how often english alphabet letters appear in the beginning of words. This info is available online in excellent article by P. Norvig.
Next, i assumed that each word is a path of a Markov process and thus information on how likely it is to get a particular letter given the current letter is required. This transition matrix can be extracted by scraping contents of this appendix of another great article.
Last piece of information is the distribution of word lengths in english language. Here i decided to take a somewhat custom approach that uses ideas from this video (very good watch btw!). In particular i downloaded a copy of Twenty Thousand Leagues Under the Seas and inferred the word-length distribution based on such corpus.
This way, to generate a random word one can draw a word length from the corresponding distribution, draw a letter from the distribution of first letters in a word and finally generate a markov chain of letters until the word of desired length is generated.
Note: for words of length 1 only two possibilites are allowed: word a and word I (like in actual english).
Use generate-random-word, generate-random-sentence, generate-random-paragraph or generate-random-text from src/textgen/main.clj.
Ri bente har heered aco. Ale an sthen desop thek. Alyon amsi alssatou ongu whav. Thaneosa rt th omit ero. Dungrle rti atwaic i ac.
Stn lc ai cubs an. Fachedv onilin ssitupa wi sefr. Reve byse ancist a heratoforerm. Bre ppa in or alitou. Qut rndense kndof het i.
Ti bi uns of rar. Heddewe anyommme risso pan angeretrst. Pltif acuseigi nthist abun alleape. Mede eng frys nstarhed rst. Lds todmere ath ts pt.