You work as a machine learning specialist for major phone and internet providers. Your customer support department needs to upgrade its phone response systems to reduce the number of human service representatives needed to handle dramatically increasing call volume. Your senior management team has leveraged off-shore call center services to reduce costs. Still, they now want to take advantage of voice recognition to automate many of the most frequent support call types, such as “I forgot my password” or “my internet is down.”
Your management team has assigned you to the team to implement the machine learning model behind the voice recognition system. Which SageMaker built-in algorithm is the best choice for this problem?

Answer options:

A.Sequence-to-Sequence
B.K-Means
C.Semantic Segmentation
D.Neural Topic Model (NTM)

Answer: A
Option A is correct. The Sequence-to-Sequence algorithm takes audio as input data and generates a sequence of tokens, such as the words in the audio. This can then be used to provide automated responses to users’ requests.
Option B is incorrect. The K-Means algorithm is used to find groups within data where the group members are similar to each other but different from members of other groups. This algorithm will not help you encode speech audio streams.
Option C is incorrect. The semantic segmentation algorithm is used to develop computer vision applications. You are trying to solve a speech recognition problem. So this algorithm would not work for this problem.
Option D is incorrect. The Neural Topic Model algorithm is used to group documents into topics using the statistical distribution of words within the documents. You are trying to solve a speech recognition problem. So this algorithm would not work for this problem.
Reference:
Please see the SageMaker developer guide titled Using Amazon SageMaker Built-in Algorithms.