This graph shows the number of appearances of three country names: Turkey, Syria, and Iraq. I used the advanced feature of setting the time period to only start in the late 1920s, after the Ottoman Empire was fully dissolved. Turkey is usually the most common, which makes sense because it is the physically largest and most populous of the three; also it was a popular vacation spot for Brittish tourists. Syria gets a small increase in the late 1940s, when it became a fully autonomous nation, and it slowly increasing in recent years due to conflict in the region. Likewise, Iraq gets a small increase in the late 1950s, when it a coup made it a sovereign nation. There is a huge increase beginning in 2003, when the U.S. invaded to find Weapons of Mass Destruction, and peaking in 2006 when war crimes became public.
Graph #2 - Gendered Pronouns
This graph shows the number of appearances of gendered pronouns in English language literatature. I used the advanced feature of setting the graph to look only at fiction, so that the equality (or lack thereof) between male and female characters can be measured. Unsurprisingly, masculine promouns were far more common than feminine pronouns. However, I was surprised that this trend persisted all the way until the mid-late 20th century - so it wasn't first-wave, but rather second-wave feminism that increased literary representation of women. One notable irregularily is that the possesive pronouns "hers" barely occurs at all. This may be due to the grammatical irregularity that there are two female possesive pronouns, one mainly used in passive voice ("She carried HER books" vs. "The books were HERS"), but only one male possessive pronoun ("He carries HIS books" and "The books were HIS").
I used Voyant Tools to analyze the same text as last week's lab, The Adventures of Huckleberry Finn. I found the contexts and the summary to be the most useful. You can see that the vocabulary desity is higher than average, because Twain introduced vernacular dialects and slang into American letters. Also, it was nice to be able to pinpoint the different contexts in which important words appear. I imagine that the trends function could be very useful, but I found it difficult to choose my intended words, and then read the resulting graph.
Sentimood
Neutral: "rise" - depends on context! "Rising grades" is good, but "Rising tides is bad.
"woke" - also depends on context. "Woke" in slang means aware of injustice, but it's also sometimes used in a joking/condescending way. To get "woke up" in the middle of the night could also be good or bad, depending on the situation.
Wrong: "frown" and "grin" are both neutral. I think "grin" should be positive because grinning is something you do when you're happy, and "frown" should be negative since frowning is something you do when you're sad.
I inserted the following section of my own creative writing into Sentimood and Global sentient:
"My mother sleeps all day and her boyfriend works all day. So me and Indigo can do pretty much whatever we want. I eat a whole box of freezer-burned ice cream sandwiches. Indigo eats a whole shaker of salt. We climb out to the fire escape and spit globs of spit at the bug-looking pedestrians. All the way down down down on the street, Mr. Ibrahim (whose job is to be the landlord and yell at us) feels something splatter on his hunched-up shoulder. He opens up his umbrella and we almost die from laughing too hard. We do archeology in my big sisters’ room and unearth tampons and menthol cigarettes and glossy photographs of naked people. We tell each other our deepest darkest dirtiest secrets. We kiss each other on the lips, to practice for boys."
And here are the Sentimood and Global Sentiment Results:
Both labs call the text vastly negative (maybe it is, although my intention was to mix the joy of childhood with something more menacing). In this way, Sentimood is more accurate, because it identified 5 positive and 6 negative (on the edge). Meanwhile, Global Sentiment called it negative with 92% certainty (not on the edge at all). Also, Global identified the character names as positive (I think that's true for Indigo and maybe untrue for Mr. Ibrahim). Sentimood identified these same names as neutrals.
Google Translate
Microsoft Bing
The sentence that I am translating is from the Turkish language novel My Name is Red by Orhan Pamuk:"O zaman söyle bana, aşk insanı aptal mı yapar yoksa sadece aptallar mı aşık olur?" In the English language version, and as I understand it, this means, "Tell me then, does love make one a fool or do only fools fall in love." Both Google Translate and Bing slightly mistranslate the word "aptal," (would could mean fool, stupid, or idiot, but in context means fool here). Also, Bing is unable to correctly change Turkish grammar into English grammar which results in a tense error. However, overall both machine translators wrote the general meaning of the sentence, just not as elloqueantly as a human translator.
Mask Machine Learning
I tried to teach my first program to tell whether I was wearing a mask. This could be useful because it could be a way to figure out if people are following making procudures in certain areas (although if it were actually applied to machines all over, it would probably be slammed as a terrible invasion of privacy). I took about 200 photos of each, with my head and mask in different positions. However, the machine can still definitely be tricked if you hold your head a certain way (maybe becuase I did not submit enough samples, or maybe because my webcam quality is legendarily crappy). Also, it only works consistently for me, not my friends.
Speech Machine Learning
I tried to teach my second program whether I was talking or being silent. This could be useful at the library, where I work, to make sure that people are being silent in silent rooms (although it would be pretty sad to replace hushing librarians with a hushing machine, in my opinion). This machine required fewer trials and was more accurate than the first machine. Potentially, this is because the absense of a thing versus the presence of a thing (noise) is easier to differentiate than the absense of one of element of a thing versus the presence of one element of a thing (the mask on the face image, the face image is there no matter what).