Final Exam Review - Part 1

pdf

School

Illinois Institute Of Technology *

*We aren’t endorsed by this school

Course

585

Subject

Computer Science

Date

Dec 6, 2023

Type

pdf

Pages

18

Report

Uploaded by MinisterWaterBuffaloPerson2295

Final Exam Review Part 1 1
Final Exam Logistics 2 As scheduled by Registrar: Date: Weds Dec 6, 2023 8:00 a.m. 10:00 a.m. Where: WH 113 - NOT regular classroom Other details: Remote proctoring and alternate arrangements are ONLY possible via IIT-Online office or CDR Students in remote sections and who do not live in the Chicago metro area should contact Charles (Chuck) Scott at IIT online. If applicable: contact CDR regarding accommodations
Final Exam Format 3 Format: 110 minutes Multiple choice, matching, or similar questions One correct answer per question No penalty for incorrect answers (no penalty for guessing) Answer on bubble sheet bring pencils! Material: Covers material before and after midterm (up to Session 27) with a focus on material after the midterm Some questions apply concepts from both 1st and 2nd half of class.
Your preview ends here
Eager to read complete document? Join bartleby learn and gain access to the full version
  • Access to all documents
  • Unlimited textbook solutions
  • 24/7 expert homework help
MULTIPLE CHOICE STRATEGIES 4
1. Read the questions carefully 2. Answer the question without looking at the options 3. Eliminate the incorrect options 4. Answer all the questions 5. Manage your time 5 LINK (from UT-Mississauga): https://www.utm.utoronto.ca/asc/media/2073/download?inline Multiple Choice Strategies
MIDTERM EXAM QUESTIONS 6
Your preview ends here
Eager to read complete document? Join bartleby learn and gain access to the full version
  • Access to all documents
  • Unlimited textbook solutions
  • 24/7 expert homework help
Multiple Choice Question 7 Which of the following describes an appropriate application of KL Divergence? Select the BEST answer. A. B. C. D.
Multiple Choice Question 8 Which of the following describes an appropriate application of KL Divergence? Select the BEST answer. A. Generating a coding system given a reference distribution of unigram words B. Evaluating a text classifier by analyzing how it will perform on unseen data, that is, how frequently the classifier’s labels match the true labels C. Comparing machine-generated texts to human-generated texts, to evaluate the quality of a machine translation system D. Outputting a prior distribution over document topic labels
Multiple Choice Question 9 Which of the following describes a difference between L1 and L2- Regularization? Select the BEST answer. A. B. C. D.
Your preview ends here
Eager to read complete document? Join bartleby learn and gain access to the full version
  • Access to all documents
  • Unlimited textbook solutions
  • 24/7 expert homework help
Multiple Choice Question 10 Which of the following describes a difference between L1 and L2- Regularization? Select the BEST answer. A. L1-regularization is intended to increase model complexity, and L2-regularization is applied to decrease model complexity. B. L2 regularization tends to push model weights to zero, removing those features from the model, and L1-regularization does not have this property. C. L1- and L2-regularization differ in the type of vector normalization applied to penalize model parameters. D. L1-regulariztion penalizes unigram weights, and L2- regularization penalizes bigram weights.
Multiple Choice Question 11 You are developing a word-sense-disambiguation model. Your classmate suggests that you evaluate your tool by generating a dataset of pseudo- words by collapsing two semantically unrelated terms into a single token (e.g., “door” and “banana” into “door - banana”). What is a disadvantage of this approach? Choose the best answer A. B. C. D.
Multiple Choice Question 12 You are developing a word-sense-disambiguation model. Your classmate suggests that you evaluate your tool by generating a dataset of pseudo- words by collapsing two semantically unrelated terms into a single token (e.g., “door” and “banana” into “door - banana”). What is a disadvantage of this approach? Choose the best answer A. This approach requires additional hand-annotated data, which is expensive and time-consuming to collect B. The resulting Cohen’s Kappa statistic is likely to be low. C. The resulting dataset may not be representative of the challenges for word sense disambiguation that occur in natural language. D. You cannot use bag of words (BOW) methods with this approach
Your preview ends here
Eager to read complete document? Join bartleby learn and gain access to the full version
  • Access to all documents
  • Unlimited textbook solutions
  • 24/7 expert homework help
Multiple Choice Question 13 You are building a bag-of-words (BOW) text classifier over English texts. Which of the following are appropriate word normalization steps that you may want to consider? Select the BEST answer A. B. C. D.
Multiple Choice Question 14 You are building a bag-of-words (BOW) text classifier over English texts. Which of the following are appropriate word normalization steps that you may want to consider? Select the BEST answer A. Using space characters as a delimiter to segment the input text into words. B. Applying an English language stemmer to remove suffixes, such as "-s" and "-ing" C. Removing words from your training set that do not appear in your test set D. Removing texts from the test set that contain out-of- vocabulary (OOV) words
Multiple Choice Question 15 Which of the following is NOT true of sentiment analysis? Select the BEST answer. A. B. C. D.
Your preview ends here
Eager to read complete document? Join bartleby learn and gain access to the full version
  • Access to all documents
  • Unlimited textbook solutions
  • 24/7 expert homework help
Multiple Choice Question 16 Which of the following is NOT true of sentiment analysis? Select the BEST answer. A. Lexicon-based methods that use positive and negative word lists have been shown to be effective for sentiment analysis. B. Word embeddings, such as word2vec, are often unable to represent positive and negative sentiment. C. Sentiment analysis is often not a strictly two-class problem and may be best expressed over a scale, or with a "neutral" label. D. A sentence may contain multiple aspects with different sentiment polarity.
Multiple Choice Question 17 You have been asked to build a named-entity recognition (NER) tool for a specific biomedical sub-domain. Your teammate suggests using TF-IDF, because this method was very successful for a document clustering task. What is an issue you may encounter with this suggestion? A. B. C. D. [NOTE: This was not a mid-term exam question]
Multiple Choice Question 18 You have been asked to build a named-entity recognition (NER) tool for a specific biomedical sub-domain. Your teammate suggests using TF-IDF, because this method was very successful for a document clustering task. What is an issue you may encounter with this suggestion? A. It may be difficult to find document-level annotation for this specific sub-domain B. TF-IDF removes word order which is important for NER C. Lexical resources such as dictionaries may not exist for this sub-domain D. All of the above are true
Your preview ends here
Eager to read complete document? Join bartleby learn and gain access to the full version
  • Access to all documents
  • Unlimited textbook solutions
  • 24/7 expert homework help