📚 ANSWER KEY 📚

• If counting contractions as single tokens (can't = 1): ~104 tokens
• If splitting contractions (can't = 2: "can" + "n't"): ~106 tokens
• If counting all punctuation marks separately: ~110 tokens

• "Tuesday," "Francisco." "incredible!" "Marcus." "ago" (could have comma) "questions." "tokens." "naturally." "$45.99" "shop." "technology!"

• can't - Contraction: Could be split into "can" + "n't" or kept as one token
• $45.99 - Number with currency symbol and decimal point
• step-by-step - Hyphenated compound word
• 50,000 - Large number with comma separator

Sample answer: Tokenization is difficult for computers because language has many special cases. Contractions need to be split or kept together, numbers can include commas and decimals, hyphenated words might be one word or multiple, and punctuation can be part of a word (like in "$45.99") or separate. Computers need rules for every possible situation, unlike humans who understand context naturally.

Sample answer: "opened" was tricky because it could be a verb (The museum opened) or an adjective (the opened box). In this sentence it's a verb. Also, "three" could be a number, but in grammar it's functioning as an adjective describing "months."

Sample answer: Yes! The word "light" can be different POS:

• "The light is bright" (noun)
• "Light the candle" (verb)
• "This box is light" (adjective)

Sample answers:

• News organizations can automatically tag articles by people, places, and events mentioned
• Search engines use NER to understand what you're looking for ("restaurants in Boston")
• Calendar apps can extract dates and times from emails to create reminders
• Business intelligence tools track mentions of companies and competitors

Sample answer: Computers face several challenges: (1) Names can be common words (Apple is both a fruit and a company), (2) Context matters to determine entity type, (3) New names and entities are created all the time, (4) Different cultures have different naming patterns, (5) Abbreviations and nicknames need to be linked to full names.

Sample answer: The passage is clearly positive because Emma is described as "amazed," she calls the technology "incredible," she finds things "fascinating," and she "can't wait" to explore more. There are no negative statements or complaints. The entire tone is enthusiastic and excited about AI and the museum visit.

• "The museum opened three months ago" - factual date
• "Emma purchased a robotics kit for $45.99" - factual transaction
• "AI systems break down sentences into smaller parts called tokens" - technical explanation
• "has already welcomed over 50,000 visitors" - statistical fact

Sample answer: If read sarcastically, "incredible" could mean the opposite - that the technology is disappointing or unimpressive. For example, if Emma said "This technology is incredible" with an eye roll, it would actually be negative. This shows why sarcasm is so hard for AI - the same words have opposite meanings depending on tone and context.

Sample answer: Computers would struggle because: (1) They can't hear tone of voice, (2) They might not understand that "can't wait" is positive enthusiasm, not literally being unable to wait, (3) Context matters - "incredible" could be positive or sarcastic, (4) Cultural differences in how emotions are expressed, (5) Subtle emotional nuances that require life experience to understand.

Sample answers:

• Companies analyze customer reviews to understand satisfaction levels
• Social media platforms monitor posts to detect harmful content or users in distress
• Political campaigns analyze public opinion about candidates
• Movie studios track reactions to trailers and films

Sample answers (students should write something similar):

• "Emma Rodriguez visits an AI museum and learns about natural language processing."
• "A student's visit to a museum teaches her about how computers understand language."
• "Emma's experience at the Artificial Intelligence Museum exploring NLP technology."

• Artificial Intelligence Museum
• Natural Language Processing (NLP)
• How computers understand human language
• Chatbots and AI systems
• Breaking sentences into tokens

Sample answer: "Emma visited an AI museum in San Francisco and learned how computers use natural language processing to understand human language."

(Exactly 20 words)

Emma Rodriguez

Sample answer: The previous sentences are all about Emma - she was amazed, she exclaimed, Emma learned. "She" comes right after talking about what Emma learned, so it must refer to Emma. Also, Emma is the only female person mentioned who would be finding things fascinating at the museum.

Sample answer: Computers have to track which nouns the pronoun "she" could refer back to. In this passage, "she" appears multiple times and could theoretically refer to any female person mentioned. Computers need complex rules to understand pronoun references (called "anaphora resolution") and must consider sentence structure, proximity, and context - things humans do automatically without thinking.

Sample answers might include:

• "Tokenization was easiest because I just had to count words and look for spaces"
• "Sentiment analysis was easiest because the passage was clearly happy and positive"
• "Named entity recognition was easiest because names and dates are obvious"

Sample answers might include:

• "POS tagging was hard because some words can be multiple parts of speech"
• "Semantic understanding was challenging because you have to track pronouns and references"
• "Key phrase extraction was difficult because everything seemed important"

Sample answer: "When I analyzed manually, I used my understanding of context, grammar rules I learned, and common sense. A computer has to follow strict algorithms and rules for every step. I can 'just know' what a word means, but a computer has to check thousands of patterns and examples. Humans are flexible; computers are systematic."

Sample answers:

• "Understanding context is the biggest challenge because words change meaning based on situation"
• "Dealing with ambiguity - humans use common sense but computers need explicit instructions"
• "Sarcasm, idioms, and figurative language don't make literal sense"
• "Language constantly evolves with new slang and expressions"

Sample answers:

• "A homework helper that can read my essay and suggest improvements would save me time"
• "Better autocorrect that understands context would reduce embarrassing text message errors"
• "Voice assistants that actually understand what I mean, not just the words I say"
• "Translation apps that work perfectly so I could talk to people in any language"

Sample answers (should demonstrate understanding of NLP concepts):

• "An app that reads news articles and creates simple summaries for kids to understand current events"
• "A chatbot that helps students who are feeling anxious by recognizing emotional words and responding supportively"
• "A tool that analyzes social media posts to detect cyberbullying before it gets worse"
• "An app that translates slang and idioms for English language learners"