920294915, Roko Pozaric

This lab was probably the most interesting one. Although it was really time consuming, I think it was worth it, and it was fun doing it. The most interesting part to me was maschine learning where I spend a lot of time experimenting and having fun after finishing my lab.

PART 1

When we put a * in place of a word, the Ngram will display the top ten substitutions.

_INF represents various grammatical categories of a word

PART 2

The novel I picked is "Moby Dick" I really liked the word cloud because I think it's very practical to use. I particulary liked stacked bar display because it's the most unusual way to represent the data ( at least to me )

PART 3
1.Positive or negative word: a) I think you have good workers. I think you do not have good workers (do not)

Every time we use negative word infront something positive, it still counts as positive.

b) promise (You broke the promise to deliver, or As promised you delivered the package)

2. Three words for which I think weighting is seriously wrong: resend, want, fell

3. They both agree:a) Hello Roko, Thank you again for providing this information and for being proactive in obtaining your immunizations so that you are compliant with our requirements. I have copied the McCosh Health Center frond desk staff on this email as they will be able to assist you in scheduling an appointment for your immunizations. Please email them directly at UHSfrontdesk@princeton.edu to schedule your immunization appointment. I look forward to meeting you soon and let me know if you have any other questions.

b) Hello Roko, I am happy to hear you have arrived safely into Princeton, and welcome! I am sorry we were not able to meet during any of the available appointment times. I am currently travelling back with my family and will not be able to meet at this point until later this week. Once I settle back in (this Wed), I can access my calendar and will send you my availability then. To not further delay your approval for enrollment, please fill out your APF and include a course queue submission as soon as possible. Best, Jon

4. They differ in their assignment: a) I am happy to inform you that your shipment is going to be late only 4 days. b) There was an earthquake last night and the building fell apart, but luckily no one got hurt.

5. They both agree but clearly are both wrong: a) Kim Jong Un died in sleep last night. b) Terrorists were very happy because they completed their mission.

Part 4
1. Two examples that are wrong: a) A third draft of the outcome of the Cop26 climate summit retained key resolutions to pursue greenhouse gas emissions cuts in line with holding global temperature rises to 1.5C.(when running it through google translate it replaces word holding with keeping which changes the complete meaning of the passage. Bing does good job translating it back from croatian to english.)
b) Young people are calculating the consequences of the climate crisis on their future, particularly with regards to childbearing. (Google translate changes the "particulary with regards to childbearing" to "especially in terms of births", which is not quite the same. Bing again does the perfect job translating it back.)
2. Two examples that work well: a) The Intergovernmental Panel on Climate Change has said emissions must be cut by 45% by 2030 compared with 2010 levels to stay within 1.5C. (when runing it through google translator, some words change but the meaning stays the same. Bing's translation back to english is perfect)
b) “What things?” I asked. Was she thinking about the heatwave in Vermont, forest fires in California or the pandemic? ( Both do the perfect job in translating this sentece back from croatian to english)
The geoogle translator and bing both do very well in translating from english to my native language. The only meaningful difference I see between google translator and bing is that bing does much better job running words from another language, back to english.

Part 5

I had to make an unequal number of image samples for class 1 and class 2 in order to get an accurate result. As my screenshot points out, it is working very well. More training examples made it less accurate.

As seen from screenshot above, this experiment worked perfectly, I wanted from program to recognize the Croatian flag and it did. In contrast to one before, I had the same nuber of learning samples. I was not aiming for 73 samples (it was an accident), and I was thinking if I had more samples, it would probably be more accurate.