wjw3, William Wallace

For the most part I enjoyed this lab and found the exercises quite useful. However, this lab took by far the most time of labs we have done, as playing around to figure out all the features of these tools took a lot of time.

Part 1

This first graph shows the frequency of the labeled terms found in books using Google Ngram Viewer. As shown below, the verb form of run and walk are much more common than their noun version, while exercise is the opposite. This graph was made using Google Ngram's advanced wildcard functionality to plot the most common parts of speech.

This next graph shows the frequency of wildcard words associated before Wallace. Interestingly, William Wallace is the currently the fifth most common combination, and was the second most common in 1915!

Part 2

Below is a word cloud of the most commonly used words in Pinocchio. As is obvious, the name Pinocchio is the most commonly used word, with puppet right after. I found the trend graphs quite interesting to see when certain words are most commonly used. I also think that the context tool (pictured below) is also very useful to find the different uses of a certain word.

Part 3

Sentimood lists the word crazy with a score of negative 2. However, crazy could also be a positive word in a sentence like: that party was crazy. The word silly is also listed with a score of negative 1, but can similarly take on positive meanings.

The word wow has a score of 4, which I think is far too high, since wow can be used with very casual meaning and is not always necessarily positive. The word fraud has a score of negative 4, which I think is far too negative for the word, as it is grouped in with far worse insults and explicit terms.

Examples where Sentimood and commercial analyzer agree and appear to be correct:
I hate you. (Score of -3, negative with 100% confidence)
I am having so much fun. (Score of 4, positive with 97% confidence)

Examples where Sentimood and commercial analyzer disagree:
I am not having fun. (Score of 4, negative with 89% confidence)
This is the least interesting thing I have ever read. (Score of 2, negative with 92% confidence)

Examples where Sentimood and commercial analyzer agree but both seem to be wrong:
I am amazed by how you are just not smart. (Score of 3, positive with 100% confidence)
My dog is so silly. (Score of -1, negative with 100% confidence)

Part 4

Below is are a few examples of translation software working well and not working well.

These examples show that Microsoft's translation for Chinese handles weird English idioms a bit better than Google does. However, in my experience neither translation software works perfectly in handling longer sentences well. Particularly, I have noticed that they have problems working with complex grammar structures, as they never translate perfectly into English.

Part 5

For this first experiment, I took a few pictures of myself and my dad, then took trained the model and took turns standing in front of the computer camera. We also tried changing the lighting and clothes, but the machine was incredibly accurate at picking who was the correct person.

The second model I uploaded pictures of Golden Retrievers and Golden Doodles as the two classes, then trained the model and input photos to be judged by the machine. Even with only a few pictures of each, the machine correctly guessed the breed in every case I gave it.