Are you sure you want to delete this post?
Again, thank you so much for your suggestions. Following I will answer various questions:
1. What information could I get from text mining?
There are various kinds of analysis, that can be conducted by using text mining. It is important to have in mind, that text mining is used on large amounts of text that can not be analyzed by reading it manually (at least not reliably and without brave research assistants). Basically one can count the number of words as a proxy for the coverage of topics. One could also use topic analyses or cooccurrence-tests to discover, which topics are covered and to get a better understanding of the framing and connotation of those topics. Advanced techniques include sentiment analysis, which promises to be able to measure the tone of a whole speech or passages mentioning certain aspects (for example policy areas, running mates or other candidates).
2. Will I implement the reactions of the crowd?
I don't think so. First, it would be very difficult, as text mining first transforms speeches into a bag-of-word. This is a well-established method and it delivers not perfect but reliable results. Measuring the reaction of the crowd would demand for a linkage of special passages within a speech with the crowds reaction. Albeit this could probably work by introducing a meta-information, it would boost the complexity of my project a lot. This brings me to my second point: I am not sure, in what dimensions the reactionsof the crowd could provide analytical variation. Even if I would be able to measure, how loud the crowd cheered, it would be no surprise, that both candidates get a lot of cheering and applause. An interesting aspect is, though, (has there to be a comma? I'm sorry...) to measure WHAT the listeners chanted exactly. Probably there would be much less, if any, hateful and negative slogans shouted by the Obama-crowds.
3. Does my method reflect the reading of the author (I understand this as you are referring to me as the author (of the thesis))?
Yes. I will deduct hypotheses from well-established theoretical frameworks on various topics. Then I conduct a systematic examination of those hypotheses. Therefore it should be transparent and reproducable how I came to my results and conclusions.
4. Will I take the context and "style" of delivery into account?
This is another important suggestion, I am most grateful for! Indeed, I could implement hypotheses about the influence of certain political incidents on the contents of the speech. The style of delivery is another very important point. For example I will measure the length of sentences, the characters of the sentence structure and the number of words used (as in vocabulary).
Do you have other aspects of "delivery" in mind?
All the best to you!