Text Mining

After having a great dinner and talk with Dr. Tao Xie I became interested with Text Mining, and read through the information that he gave me. Text mining is a very useful tool that I believe will be in high demand in the near future. The ability to summarize and group large amounts of text such as emails, customer complaints, and surveys is very valuable to small businesses to big corporations.

SAS Text Miner gives you the capability to perform both descriptive and predictive text mining. From my understanding SAS does this by first processing the data using %TMFILER macro which creates a SAS data set. The data set is used in the text ming node for text parsing and then transformed into a informative format. Then it is analyzed for predictive and descriptive purposes.

Text Mining is a very powerful tool and is well on its way to become a high demand software. The %TMFILTER macro is so powerful and useful especially with the support of several languages, the possibilities are endless.

Travis W.

End of the sprint...

IBM has incorporated agile into their design structure and for me, this is nearing the end of this sprint. It has proved to be a very rewarding experience here at IBM.

That being said. I have been overloaded with work here and have found myself little time to work on research. I am hoping that the end of this sprint will allow me sometime to focus on returning to working on the research.

Itutor Update

The past two weeks made it very difficult for me to work on the iTutor eclipse plugin, because of my recent move from my old apartment. Limiting my accessibility to many of my necessary resources such as the internet, and not being in Raleigh. Now I am situated and continuing working on the plugin.

The previous version of the iTutor plugin decompiled the teacher's solution (.class) and the students solution (.class) and displayed method information of the decompiled classes. Currently I have been working on changing the GUI to read in the source directory and path directory of the TA solution and Student solution as requested by Jcute in order to create and run regression test on both solutions and display where the two solutions differ, and give the student an idea where they went wrong and what they need to correct.

Might reach conflict with previous code when requesting for the solutions, for the simple fact that JCute requires (.java) files and the previous code request for the (.class) files. Which I believe is a simple fix.

Co-op and GUI

Since last week was my first week at IBM co-oping, I was not able to put in many hours. I have also decided to move my updates to Sunday. This will better allow me to make meaningful updates.

Travis has made a class that will actually run jCute. The problem is, it seems to be hard coded and we need a flexible way to modify input files. I have been looking at plugins with eclipse and have spent about 4 hours trying to create a GUI interface for what Travis has worked on. I hope to have it submitted this Saturday into the SVN so there is a deliverable for work accomplished.

BERT

For the second part of my summer research I began working with BERT. So i put a pause on implementing JCute into iTutor. BERT is an eclipse plugin for regression testing. Bert allows you to select several projects in your workspace and checks behavioral differences in new versions of your project. This could be a very good tool to implement with iTutor when comparing the teachers "Golden Solution" with a students solution. The BERT view also displays the logs of version comparisons winch will also be very useful.

BERT also brought about some adversity. The plug-in was written on a Linux machine, and the programmers did not make the plug-in Windows friendly. As instructed by Dr.Xie I began trying to debug the code. This was a very daunting task that I didn't spend too much time on. we decided that it would be better for me to continue working with JCute.

iTtutor JCute Integration

I have been working on running the JCute batch file in iTutor for testing the given programs path coverage. I hard coded the batch file and ran it using a command runner class, and was successful printing out the path coverage information in the console.

The next step in the process will be creating a batch file with the users choice of options, location of the source directory, and class file. Currently I plan to create the Batch.java file then run it using the command prompt. The most difficult part in this task is creating the Batch file, and being able to display results in the iTutor perspective.I later plan to alter the iTutor properties, in order for it to include JCute options.

Linux/Windows

For the last week I have been trying to modify an existing code to work under windows. Now the process itself is not too difficult if the programmers of the original program intended for the program to be cross platform. That way the inherent program will leave room for a different operating system and commands.

Now the norm with most code is that there is a timetable required for deliverables. I would assume the same issue was with the existing code. The portion that I wanted to make working within windows had the top half hard coding some values and the second half using a more robust system. I would imagine there would have been two people working on this and trying to has it together to get it working.

I have mostly given up on trying to do the work of integrating their system with windows. If I have time I will take a look at it and modify it until I can get it working, but there seem to be a lot of underlying work.

Since I myself have a timetable to work with, I am moving my focus to working on iTutor itself. I am hoping that I will get some tangible work done so that there is a deliverable.