Remove overpresent / bias words from extracted text

[lea-33]

related to #47

Create wordclouds for all slides and remove the most occuring und unneccessary words (names, filling words, single letters,..). Do that iteratively (2-3 times) and then calculate the word embedding again and check if it made a difference.