Approaches to Data Labelling CT-AI Practice Questions (Page 6)

Questions 26–30

1. 26foundation · medium

   What is the primary purpose of data labelling in supervised machine learning?

   • ATo reduce the size of the dataset by removing irrelevant features
   • BTo encrypt the data so that it is secure during model training
   • CTo automatically generate new training examples from existing ones
   • DTo assign meaningful target values to raw data so the model can learn the relationship between inputs and outputs
   Reveal Solution Discussion Select an answer first
2. 27foundation · medium

   Which of the following is a key feature of a labelling tool that supports quality assurance?

   • AThe ability to automatically deploy models to production
   • BThe ability to track annotator performance and review labels
   • CThe ability to generate synthetic data without human input
   • DThe ability to run the model on GPUs for faster inference
   Reveal Solution Discussion Select an answer first
4. 28foundation · medium

   Which labelling approach is commonly used for audio data?

   • ADrawing bounding boxes around objects
   • BHighlighting text spans for named entities
   • CTagging entire images with a single class label
   • DTranscribing speech to text or annotating time segments with labels
   Reveal Solution Discussion Select an answer first
5. 29expert · hard

   A company is labelling a dataset of voice recordings for a speech recognition model. The recordings contain personal conversations. They plan to use a crowdsourcing platform. What is the most important ethical and privacy consideration, and how should they address it?

   • APay crowd workers a higher rate to ensure they handle the data responsibly.
   • BProvide clear labelling guidelines to the crowd workers to ensure accurate transcription.
   • CUse a model-based automated labelling system to avoid exposing the recordings to humans.
   • DObtain consent from the speakers and de-identify the recordings before sharing them with crowd workers.
   Reveal Solution Discussion Select an answer first
6. 30foundation · medium

   What is a key advantage of manual labelling over automated labelling?

   • AIt guarantees perfect consistency across all labels
   • BIt is always faster for large datasets
   • CIt can handle complex or ambiguous cases that automated rules may miss
   • DIt is less expensive for large-scale projects
   Reveal Solution Discussion Select an answer first