Text Analytics DATA-SCIENCE-FOUNDATIONS Practice Questions (Page 5)

Questions 21–25

1. 21foundation · easy

   What is the primary purpose of text analytics in a business context?

   • ATo extract meaningful patterns and insights from unstructured text data
   • BTo store large volumes of text data in a data warehouse
   • CTo replace all human decision-making with automated text generation
   • DTo convert structured data into unstructured text for storage efficiency
   Reveal Solution Discussion Select an answer first
2. 22foundation · easy

   In text preprocessing, what is the purpose of removing stop words?

   • ATo convert all words to their base form for easier comparison
   • BTo identify the most important keywords in a document
   • CTo reduce the dimensionality of the text data by eliminating common words that carry little semantic weight
   • DTo replace rare words with more common synonyms
   Reveal Solution Discussion Select an answer first
4. 23application · medium

   A customer support team wants to automatically flag negative feedback from support tickets. They have a labeled dataset of 5,000 tickets with positive and negative sentiment labels. The team needs a model that can learn complex language patterns, including negations like 'not good'. Which approach should they use?

   • AUse a simple keyword matching script that flags tickets containing 'bad' or 'poor'
   • BPerform topic modeling with LDA to identify negative topics
   • CApply a lexicon-based sentiment analysis using a predefined word list
   • DTrain a machine learning classifier on TF-IDF features
   Reveal Solution Discussion Select an answer first
5. 24expert · hard

   A company wants to analyze sentiment in customer reviews of a new smartphone. The reviews contain sarcastic phrases like 'Great, another update that breaks my battery life.' A lexicon-based sentiment analysis tool classifies this as positive because of the word 'great'. The team needs a more accurate approach. Which strategy should they use?

   • AAdd more positive words to the lexicon to improve recall
   • BUse topic modeling to identify sarcastic topics
   • CRemove all stop words before applying the lexicon
   • DTrain a supervised machine learning classifier on a labeled dataset that includes sarcastic examples
   Reveal Solution Discussion Select an answer first
6. 25expert · hard

   A legal firm is building a text classification model to categorize contracts into 'non-disclosure', 'employment', or 'partnership' types. They have a labeled dataset of 2,000 contracts. The dataset is imbalanced: 80% are non-disclosure agreements, 10% employment, and 10% partnership. The model achieves 90% accuracy but classifies almost all contracts as non-disclosure. Which evaluation metric should they use to better assess model performance?

   • APrecision only, because it measures how many predicted labels are correct
   • BAccuracy, because it is the most intuitive and widely understood metric
   • CF1-score, because it balances precision and recall for each class, especially for minority classes
   • DRecall only, because it measures how many true labels are correctly predicted
   Reveal Solution Discussion Select an answer first