Mining the Elections
Our group can use text mining to answer the historical question that the group has proposed about if and how the outcome of presidential debates determined who won the election. Text mining would allow us to see if there were key words or phrases used by candidates during the debates that proved to have a positive or negative effect on voters, and as result, attracted voters or deterred them away. Another way text mining will be beneficial in our project is to determine if other aspects apart from analytics played a role in deciding the outcome of elections based on a candidate’s performance during the debates. During debates, candidates present various types of data to present their case to voters: statistical data, such as their previous track record while serving in their current governmental post; and conditional data, such as what they expect to accomplish if they are chosen as president. Because debating not only deals with factual data presented by the candidates but also the manner in which the candidates convey the data, such as their behavioral disposition, body language, tone of voice, eye contact, etc., data mining will help capture the effects of these different factors and what role they played in steering the outcome of the election. However, we keep in mind that our analysis is on the premise that the election process is very complex and trying to keep all variables stable poses multifaceted challenges.
Very good point, Felipe. The data sets you describe are certainly large enough that mining can lead to conclusions unreachable without computer aid. Although in order to explain the type of causation you are asking about (e.g., how did the debates, either individually or collectively, sway voters?) you will need to conduct some very sophisticated text mining carefully calibrated to measure degrees of change. To choose a trickier example, what formula would you apply when assessing the impact of body language on voters? What data would you input? What type of metrics would you expect to output?
Those are very interesting questions Tom, particularly regarding the formula to use in assessing the impact of body language on voters. I believe to answer that question we need to ask what are the important and calculable variables needed in combination to answer that specific question. Fortunately with the perpetual advancement in technology we have some very new and savvy tools to help us answer these questions.
There are new technology where a group of homogenous voters could use a dial in realtime to calculate the responsiveness to the candidates as they debate. This could be very helpful in identifying body language as well as candidates’ responses to the various questions posed to them. I understand that using this new technology has some intricate issues to be worked out; such as the meaning of say, a lack of eye contact with one voter to another. However, I believe that prior to any debate there are rules and customs of the culture that if not followed could prove negative to candidates who fail to oblige. Another method of gauging body language could be the use of social media such as tumbler, twitter and Facebook; which could be used to glean and identify developing themes about things such as their reaction to the candidates’ body language.
Obviously collecting data as mentioned above would be difficult, if not impossible to process and calculate to make sweeping statements about the elections. However, our goal isn’t to make sweeping statements but to identify the correlation of the debates on the presidential elections. To do so, we’ll probably need to identify the different types of data needed to answer a specific question. Like say, response to substance or correctness of facts of an answer as opposed to body language and things of that nature.
As for the metrics I expect to output, I have no idea what you are talking about. No! Just joking! I think we can expect different output for the specific questions that might vary with regards to the method of assessing the question.