email for 10/7/14's work session

agency-10-7-14-worksession-email.md

Hey everyone!

Tonight's another Agency work session night! Unfortunately, several of our officers will be gone for tests/school events, so we're going to have a bit of a different approach tonight.

First off, if anyone did any work over the past week - discuss it amongst yourselves - tell what you did, what you found, etc. Additionally, if anyone would like to put this work up on the Agency's github, at the end of this message is a set of instructions for doing so.

Now, as far as the meeting: I've settled on some tasks that hopefully you all can accomplish during today's meeting. Work together! Share screens, communicate. I'll do my best to stop by as soon as my test is over if you guys are still working. All of this is flexible - I'm not in charge here - this is y'alls project! Feel free to change it up, do something different - this is just a set of suggestions for how I think we should progress.

Remember, there is no meeting next Tuesday due to Fall Break, so let's try to get a lot done before the meeting after that!

#Data Collection

1. Find a decently sized set of headshots of celebrities/politicians/famous people. E.g, IMDB images of actors, whitehouse.gov images of presidents, etc.
   • Must be the same size
   • Ideally taken under similar circumstances (e.g, similar lighting, same background, etc.)
2. For each person/headshot, find a speech that they gave
   • Ideally given under similar circumstances (e.g, presidential inaugration speeches, Academy Award acceptance speeches, etc.)
   • Ideally similar lengths
3. For each person/headshot/speech, identify their Myers-Briggs personality type
   • four-letter type, e.g INTJ, ESFP, etc.
   • if you can't find this, find general personality traits

#Data Processing

Here you can still work together, but eventually you'll need to accumulate all this data together on one person's computer.

1. Standardize all the image names and put them all together in a folder. E.g, "FirstNameLastName.png"
2. Copy the text of each speech into a text file and once again standardize the name/format - e.g, "FirstNameLastName.txt" a. be consistent with your image names
3. Copy the personality types into some sort of file and again standardize the name/format, e.g "FirstNameLastNameTYPE.txt" a. If you couldn't find Myers-Briggs four letter types, just be sure to standardize what you did find.

At this point, if somebody in the group wants to attempt to put the data on Github under the whoareyou project (preferably in a new folder called "data" or something) go ahead! If you'd rather not, just send it to one of our officers and we'll get it up there.

#Running NLP & CV, ML This is less set in stone, but:

1. Try using Python NLTK (see the code I've put on Github as an example) to experiment with parsing the speeches. a. can you identify any patterns amongst people's speeches?
2. Use something like NumPY in Python to take eigenfaces of the pictures (see https://github.com/brigr/py-eigenfaces/blob/master/eigenfaces.py as an example) a. again, can you identify any patterns amongst eigenfaces?
3. Machine Learning people - consider how you could train a supervised learning agent on speeches/faces to recognize personality type. Work with the NLP and CV team to get data in the format you want!

#Git instructions

1. Visit: https://github.com/gtagency/whoareyou
2. In the upper right, click "Fork". This will create a copy of the repository in your account.
3. Clone this repository to your computer using git clone https://github.com/YOURUSER/whoareyou
4. Make the changes you want to make
5. Add the files you want to track with git add filename. For example, if you added a "data" folder with all the data inside, you could run git add data
6. Once you've added the files you want to commit, run git commit -m "description of what you did here"
7. Then, git push origin master, type your Github username and password, to push this back to your github repo.
8. Once you're ready to have one of us confirm your changes, click the "pull requests" tab on the right and create a new pull request. Add a description and the update details, and send it off! As soon as we can look over it we should merge it back into the Agency's repository.