Week 1

Upon arrival in the lab on the first day, was introduced to the project designated as Databuddies. An explanation of the project that would be occurring for the summer. The project is in cooperation with CRA to design a visualization method for the surveys conducted by the group. After being designated with the task of free text sentiment analysis, I began to search through articles about the subjects of topic modelling, and sentiment analysis for a Lit Review.

Week 2

After a Fair amount in research through papers, I met with Dr. Metoyer about a strategy to analyse free text data. There was a discussion of SentiWordNet, a java program built by researchers that compiles a database of multiple opinion scores of positive and negative and attaches the score to a word. My next task was to learn how to use this program effectively, so I began to compile codes of java in order to get the program working, requiring the relearning of alanguage I had not used in years. Once I successfully was able to use the API and source code, I was able to present my findings and plotted a plan of action with Dr. Metoyer.

Week 3

After discussing the goals for the sentiment analysis program with Dr. Metoyer, I began working on a program that would be able to take an entire block of text and be able to attach a sentiment score to each non objective word. This made me realize I needed a Parts of Speech tagger, to ensure the proper form of the word was sent into SentiWordNet. So with the help of a POS tagger known as LBJ POS, I implemented this library into my program. During this week also, I helped Dr. Metoyer run a workshop on 3D graphic design, to a group of middle schoolers, and rising high schoolers.

Week 4

Building the program led to difficulties in building the sentiment analysis program, based off simple quirks in the Egnlish language. One of those came down to negation, sentences were the intended sentiment score is not the same as the composite sentiment score based off of the words. An example of this would be the statement," Not Good." Though Good has a positive sentiment score, and not has a negative, their composite score ends up with the sentence being positive. Seeing as not, negates the positivity of Good, it should be negative. Another issue I worked on was the plurals of words, seeing as SentiWordNet does not recognize the plural form of words, I had to build a function into the code that will change any plural word to be processed as its singular form, returning the appropriate score.

Week 5

Debugging the sentiment analysis led me to finding a lot more issues lying deep in the program. The fifth week in the lab encompassed me spending a lot of my time picking through the several issues that turned up in the program. The bugs weren't compiling errors, instead they were errors that usability. The issues came from how we wanted to use the score from the program, which caused us to figure out which sentiment score is better and how we wanted to handle things such as adverbs.

Week 6

The resolution for the problems that arose from debugging were not solved within the first week. The nature of these issues had become more a compiling and algorithm issues because of how the sentiment analysis worked. So much of the week was spent trying to devise a better way to handle sentiment analysis. Once this was handled, I was able to make the algorithm a bit more efficient.

Week 7

This week had one focus, and that was integrating sentiment analysis into the website for CRA. The initial thought was just to use a bridge that connected java to php. Since the code was coded in java, and most of the website is done in php/html, then this was figured to be the best approach. This week was was spent learning about java servlets and how they work

Week 8

After learning a bit about java servlets, and tomcat servers while struggling to figure an optimal way for this to work, this week was spent exploring possible better options. The idea thought was to use an applet, and integrate with the website where the applet would take the information and run it using the browser's java, and the browser. This was put on the back burner because there were questions about security of the information that would be pushed.

Week 9

After sometime, the idea of the bridge was brought up again. This idea was revisited, and I took time to try and work through the bridge slowly. I was able to get the use of the servlet, and even produce a program in java that worked and did what we asked it to do. The issue came with packaging all the necessary files in to an executable WAR (web application archive). The tomcat server did not unpackage the file properly, and so it wouldn't run.

Week 10

This last week was spent cleaning up code, and trying another way for the java to work. The code included writing up readme, and making sure comments were in it to make it work. These comments and readme would give anyone the opportunity who came behind me to be able to just jump in get it to work. The next way that was attempted to fix the integration issue was to use Proc_open() command code that works with the actual local server, running the java on the server and then taking the output and putting it to the website.