Approaches to Data Labelling CT-AI Practice Questions (Page 2)

Questions 6–10

1. 6application · medium

   A company is building a model to classify legal documents into categories (contract, court ruling, etc.). They have a small team of legal experts who can label documents, but the volume is large. They decide to use a semi-automated approach where a rule-based system pre-labels documents based on keywords, and the experts review and correct the labels. What is the main advantage of this approach?

   • AIt reduces the experts' workload by pre-labelling obvious cases, allowing them to focus on ambiguous documents.
   • BIt eliminates the need for human review, saving time and money.
   • CIt ensures that the labelling process is free from bias.
   • DIt guarantees that the labels are more accurate than manual labelling.
   Reveal Solution Discussion Select an answer first
2. 7foundation · medium

   Which type of data labelling is used when the target variable is a continuous numeric value, such as predicting house prices?

   • AClassification
   • BObject detection
   • CSegmentation
   • DRegression
   Reveal Solution Discussion Select an answer first
4. 8foundation · medium

   Which practice is most effective for improving labelling consistency?

   • AAllowing each annotator to label data in their own way without oversight
   • BUsing only one annotator for the entire dataset
   • CAvoiding any form of quality checks to speed up the process
   • DProviding detailed labelling guidelines and examples to annotators
   Reveal Solution Discussion Select an answer first
5. 9application · easy

   A data science team is preparing a dataset for a supervised learning model that predicts house prices. They have collected features such as square footage, number of bedrooms, and location. What is the primary purpose of labelling this dataset?

   • ATo reduce the dimensionality of the dataset.
   • BTo identify which features are most important for the prediction.
   • CTo provide the target values (house prices) that the model will learn to predict.
   • DTo clean the data by removing outliers and missing values.
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   A non-profit organization is labelling images of wildlife from camera traps to train a model that counts animal species. They have a team of volunteers who are not experts in wildlife. To ensure the labels are reliable, what should they do?

   • AHave each image labelled by a single volunteer to maximize the number of images labelled.
   • BUse a pre-trained model to label all images and have volunteers only review the model's predictions.
   • CProvide a detailed labelling guide with examples of each species and require volunteers to complete a training quiz before labelling.
   • DLet volunteers label freely and then use a majority vote to decide the final label.
   Reveal Solution Discussion Select an answer first