Let Michael Scott be a part of your Telegram chat - see https://en.wikipedia.org/wiki/Michael_Scott_(The_Office)#Interests .
This is my very first mini-project logistic regression. Inspired by the paper http://www.aclweb.org/anthology/P11-2016 I wrote a little classifier to detect sentences with double entendres. Wrapped as a Telegram bot to make this whole thing more fun :)
Test it yourself with @twss_bot in Telegram.
My training data is based on ~7500 positive and ~7500 negative samples. Since I'm not sure about the legal part, I will not publish the training data nor my crawlers to gather it. But the trained model is part of the repository.