MLS-C01 Question 308
Select 3A company is building a customer service chatbot using Amazon Lex. The chatbot should be able to handle multiple intents, such as 'Check Account Balance' and 'Reset Password.' However, during testing, the developers notice that the bot is incorrectly identifying user intents, especially when users provide incomplete or ambiguous inputs. What actions can the developers take to improve intent recognition in Amazon Lex?
- A
Add more sample utterances for each intent to improve training data.
- B
Enable sentiment analysis in Amazon Lex to better understand user input.
- C
Use slot types with more examples or custom slot types to clarify ambiguous inputs.
- D
Optimize the confidence score threshold for intent classification in the Amazon Lex console.
- E
Integrate Amazon Comprehend to preprocess user input before passing it to Amazon Lex.
Show answer and explanation
Correct answers: A, C, D
Explanation
To improve intent recognition in Amazon Lex, developers should focus on enhancing the training data by adding more utterances and refining slot types. Adjusting the confidence score threshold can also help fine-tune intent classification. These steps directly address the issue of misclassification, while features like sentiment analysis or external tools like Amazon Comprehend are not designed for this specific purpose.
- A. Correct.
Adding more sample utterances for each intent improves the training data, allowing Amazon Lex to better understand variations in user input. This is an effective way to improve intent recognition.
- B. Incorrect.
Sentiment analysis in Amazon Lex helps understand user sentiment (e.g., positive or negative tone) but does not directly improve intent recognition.
- C. Correct.
Using slot types with more examples or custom slot types helps Amazon Lex better resolve ambiguities in slot values, which can indirectly improve intent recognition.
- D. Correct.
Optimizing the confidence score threshold allows you to fine-tune when Amazon Lex identifies an intent versus falling back to a 'no match' response, helping to reduce misclassifications.
- E. Incorrect.
Amazon Comprehend is not directly integrated with Amazon Lex for intent recognition. While useful for other preprocessing tasks, it does not address the specific issue of intent misclassification in Amazon Lex.