Microsoft Certified Azure AI Fundamentals · Free Practice Question Medium
Question 12
A travel booking chatbot frequently misinterprets customer requests. The development team wants to improve its understanding of user intentions. Which sequence of steps would BEST address this issue?
-
A
Label utterances → Deploy model → Test accuracy → Author intents
-
B
Train model → Author intents → Label utterances → Deploy model
-
C
Deploy model → Test accuracy → Author intents → Label utterances
-
D
Author intents → Label utterances → Train model → Test accuracy
Reveal correct answer
Correct answer: D
Explanation
When a travel booking chatbot misinterprets customer requests, improving its understanding of user intentions involves refining the training process. The best approach is to begin by clearly defining the conversation intents, then labeling sample utterances that reflect these intents, followed by training a model using this annotated data, and finally testing the model’s accuracy to ensure it meets the desired performance.
Step-by-Step Process for Improving Chatbot Understanding:
Authoring Intents:
The first step is to define a comprehensive set of intents that cover the various ways users may express their needs in the travel booking domain. This step lays the foundation for understanding what each user query is trying to accomplish.Labeling Utterances:
Once the intents are defined, the next step is to label a diverse set of utterances that exemplify each intent. This annotated data serves as the training set for the model and is crucial for teaching the system how to differentiate between similar expressions.Training the Model:
With a well-labeled dataset, the model can be trained to recognize patterns associated with each intent. This training process is iterative and may involve fine-tuning the model to optimize its performance based on the annotated data.Testing Accuracy:
After training, the model’s performance should be evaluated using a separate set of data to ensure that it accurately interprets user requests. Testing allows the development team to identify areas for improvement and make necessary adjustments before deploying the model into production.
Why This Sequence is Effective:
Starting with the definition of intents and the subsequent labeling of utterances ensures that the model has a clear understanding of the customer’s language and the context in which requests are made.
Training the model on this high-quality labeled data builds a robust system that can generalize well to unseen queries.
Finally, testing and evaluating the model’s performance provides the necessary feedback to iterate and improve the system, ensuring that it consistently understands user intentions correctly.
Further Reading and Documentation:
For insights into designing and training natural language understanding models, refer to the Azure AI Language service documentation, which outlines best practices for intent classification and entity recognition.
Microsoft's Educational Content Around This Topic For The AI-900 Exam:
A.
Labeling utterances before defining clear intents can lead to inconsistent or incorrect labels. Moreover, deploying the model before finalizing the intents and training it thoroughly will likely perpetuate the misinterpretations.
B.
Training the model before defining intents and labeling utterances means that the model is not being guided by a clear understanding of customer requests. This order would result in a poorly aligned model and does not solve the issue of misinterpretation.
C.
Deploying the model before defining the intents and labeling utterances means the model will be built on an undefined or incomplete dataset. This approach does not systematically address the root cause of misinterpretation, which is the lack of clear intent definitions and training data.
D.
This sequence starts by defining what the chatbot should understand (authoring intents), then collecting and labeling sample utterances to create a robust training dataset. The model is trained on this curated data, and its performance is validated through testing accuracy. This methodical approach directly addresses the issue of misinterpreting customer requests by ensuring the model is aligned with clear, well-defined intents.
Discussion
Think the marked answer is wrong, or have a better explanation? Share it below — comments appear after review.
