Kayla Hopple
One problem I had while developing this website was trying to make the background appear throughout the entire webpage without any white bars sticking out of it. I tried various CSS codes to adjust the size and position of the image, but it always led to white bars sticking out of the top and bottom no matter what. I then turned to ChatGPT for a solution, and it told me to adjust the margins in the body. I even had some problems making sure the text fit in the buttons, especially when I tried to adjust the size of both the buttons and the text, which I also contacted ChatGPT about, and it told me to adjust the size of the text and the padding, height, and width of the buttons to fit the text in the buttons.
Another problem was trying to make the word cloud have more relevant data. I tried to look for ways I could get rid of unnecessary words like "I," "an," "the," or "of," among others by trying to see if I could remove them within Antconc, but I could not find a way and then tried looking up ways to make it work. Using what I learned from there, I tried to find the suggestions within Antconc but had difficulty looking for them despite following the directions. Finally, I decided to make a whole new XML text corpus with the mentioned words removed with the "Find and Replace" feature by searching for each word and replacing it with a space.
What I thought we could do differently, other than making the AI work properly, is provide more visuals about data from different games, especially word clouds about the most frequent words. Other than that, we thought our website looked pretty well-made.

Alyssa Hopple
As we worked on our project together, I faced some hardships with some of the tasks required for the completion of our semester project. As an assistant programmer, the hardest thing I did was use Regex to convert TXT to XML after the text had been extracted from the HTML pages of the ACE ATTORNEY Wiki. I had tried to use different kinds of Regex code, including some new Regex codes we have never learned in this class, to try to wrap around each line, marked by a specific character who said that line. Many parts of the TXT extracted from the Wiki includes "i" and "Leads to" elements, which I had a hard time getting rid of. With the assistance of Reece, as well as our professor, I have gotten less bugs as I utilized the right Regex codes for this portion of the semester project.
Another hardship I faced is tinkering with the PyVis used to highlight frequent terms said by all characters, which was a Jupyter notebook originally written by our professor. I attempted to adapt that Jupyter notebook into our research, but I received undesirable programming results. However, Reece decided to take the helm and come up with a different graph with a similar evaluation, which is the most frequent terms said by Phoenix Wright himself.

Reece Cullen
For me, the biggest obstacle ended up being training the TensorFlow model to generate text. It involved so much fenagling with the example code to get it to work with our data, and then so long to train, that I kind of gave up on improving that model to try and generate something cohesive. I don't think that's a big loss, however, as the larger problem (I believe at least) is in our data. Everything is a bit too same-y, so when I would try to make the model less creative, it would just generate a bunch of super common words, like 'the', 'and', 'for', nothing unique to the Ace Attorney franchise. I wish I had chronicled all the generations I did more!
Something I ended up doing for one of our visualizations was creating a sub-folder in our text corpus containing only the chapters of the first game. An interesting avenue we could've followed would be to create similar folders for all the games, and done analysis on each game separately then compared our results. However, this would've been an arduous process, not even I, a huge Ace Attorney fan, could do this without checking back and forth with the wiki over and over again. I didn't want to do this as it would've felt unnatural to me. If we had more time and a larger visualization/analysis quota I definitely would've done it.