I don’t want to get into details on Markov chains, but in very short I would summarize it like this: For each word we collect each following word in any sentence. I decided to use a prefix or ‘lead’ of one word. After loading my text file and doing some coarse cleanup in eliminating some recurring news outlet signatures and double punctuation etc., I could finally start working the text into my Markov model. I decided that that is sufficient for such a highly unscientific and nonsensical project. After about 1 1/2 weeks I had more than 1000 headlines. All I had to do now is be patient and wait until a significant amount of headlines was accumulated.
Sentence new word generator windows#
Thankfully it is easy to set up a task scheduler in Windows which would execute my file every day in the background and let my text file grow continuously. Not bad, but of course doing it once gives very little variety of words to work with and my Markov chain is running the risk of falling to often into a unique chain of words - and in the worst case - simply repeating and already existing news headline. On any day, this would give me about 100–120 news headlines. I found this website extremely helpful to set up my little piece of code: I then isolated the text containing the headlines and wrote them into a text file. First, I set up a script which is accessing a popular news site using the keywords ‘Donald Trump’. Simply because I wanted to practice some pandas functionalities and I am not ultimately interested in the most “elegant” way to do it in this case, but I wanted a quick result and learn something on the way. Even though the Python dictionary data type seemed like an initial good fit for a Markov model, I wanted to use data frames and pandas for it.
Sentence new word generator generator#
The script consists of a quick web scraper to get as many news headlines as possible and use them in a Markov model sentence generator to create my very own ‘real fake news’ headline.
0 kommentar(er)
