Implement Custom Language Models AI-102 Practice Questions (Page 5)

Questions 21–25

1. 21foundation · easy

   Why would you export a knowledge base in Azure AI Language question answering?

   • ATo improve the accuracy of the answers
   • BTo publish the knowledge base
   • CTo train the knowledge base
   • DTo create a backup or migrate the knowledge base to another project
   Reveal Solution Discussion Select an answer first
2. 22expert · medium

   A custom language model for a healthcare appointment system has high accuracy but the team wants to reduce the number of utterances in the training data to speed up training and reduce costs. They have 10,000 utterances across 50 intents. What is the best way to reduce the dataset while maintaining accuracy?

   • ARemove all utterances that contain entities.
   • BUse active learning to identify and remove utterances that are similar to others, keeping only diverse examples.
   • CRandomly remove 50% of the utterances from each intent.
   • DReduce the number of intents by merging similar ones.
   Reveal Solution Discussion Select an answer first
4. 23application · medium

   A custom question answering knowledge base has been trained. The team wants to verify that it provides correct answers to a set of common questions before going live. What should they do?

   • AExport the knowledge base and manually review the QnA pairs.
   • BUse the test feature in the Custom Question Answering portal to submit sample questions and review the responses.
   • CRetrain the knowledge base multiple times to improve accuracy.
   • DPublish the knowledge base and ask users to report any incorrect answers.
   Reveal Solution Discussion Select an answer first
5. 24foundation · easy

   What is the purpose of creating a custom question answering project in Azure AI Language?

   • ATo train a language model for intent recognition
   • BTo translate text between languages
   • CTo generate natural language text from prompts
   • DTo build a knowledge base that can answer specific queries based on custom data
   Reveal Solution Discussion Select an answer first
6. 25expert · medium

   A company has trained a custom translator model for English-to-German technical documentation. The model performs well on the training data but poorly on new documents that contain recently introduced terminology. The team has a small set of new parallel sentences for the new terminology. What is the most efficient way to improve the model?

   • APublish the current model and use post-editing on the output.
   • BCreate a new custom translator model from scratch with the new data.
   • CAdd the new parallel sentences to the training data and retrain the custom translator model.
   • DUse the standard Translator service with a glossary for the new terminology.
   Reveal Solution Discussion Select an answer first