Gender Guesser calculates the gender of the author of any text that you try to feed her with an approximate accuracy
of 60% -70%. Results are given as a percentage of male-female and may vary depending on the style of writing - formal or spoken. To get more accurate results, use text with more than 300 words, and of course the text should be in English. The results of “weak male” or “weak female” may mean that the author writes in European English. It is very interesting to check the texts of various writers: when a writer is a man and writes on behalf of the heroine of a book or when a woman is a writer who writes under a male pseudonym. Well, it's also interesting to check your lyrics.
The program is written entirely in JavaScript - the curious can dig. For calculations, the Bayesian network is used, which estimates the frequency with which a word is used more inherent to women or men. You can read a
scientific report , detailing the details of the algorithm (in English, 153 kb). The algorithm was developed by researchers from the Institutes of the Illinois Institute of Technology (USA) and Bar-Ilan University (Israel).
PS: I will brag - a professor from the Illinois Institute of Technology, Shlomo Argamon, my current teacher in artificial intelligence.