Exam AWS Certified Machine Learning - Specialty topic 1 question 151 discussion

I will go with A. Refer to link : https://aws.amazon.com/comprehend/features/

D is the answer. Using Apache MXNet rules out Comprehend from making the classification task

Preprocessing using Comprehend. Then use preprocessed text as input to custom classifier.

Amazon Comprehend cant bring your own model, the feature of "custom classification" is meaning that your can train a classifier on the service with your own data, not bring your own model on. So the answer is definitely D.

Amazon Comprehend is a natural language processing (NLP) service that can perform part-of-speech tagging and key phrase extraction tasks. AWS Deep Learning Containers are Docker images that are pre-installed with popular deep learning frameworks such as Apache MXNet. Amazon SageMaker is a fully managed service that can help build, train, and deploy machine learning models. Using Amazon Comprehend for the text preprocessing tasks and AWS Deep Learning Containers with Amazon SageMaker to build the custom classifier is the solution that can be built most quickly to meet the requirements. References: Amazon Comprehend AWS Deep Learning Container

The Custom classification in AWS Comprehend cannot choose algorithm, you cannot use your own algorithm in it. You only feed dataset to it. So A is wrong. The data science team want to use their own MXNET model, so D.

Will go with A

A is the most quickly solution.

We have to solve two NLP problems: part-of-speech tagging and key phase extraction. Note that the custom classifier already exists and has been trained! The question asks that it be done as quickly as possible, so the idea is to use a ready-made service. Letter A is wrong, as it uses another service compared to the already created model to classify. Letter B requires development and therefore would not be the fastest solution. Letter C is wrong for the same reason as Letter A, in addition it proposes an unsupervised service (LDA) for a supervised problem. Letter D is correct.

Therefore, option D is the most efficient solution for building a NLP application that meets the requirements of the data science team.

Quickest A

The other mxnet model is the key

option D is the most appropriate answer, given that the team has already written and trained a custom classification algorithm using Apache MXNet. Option D allows the team to use Amazon Comprehend for part-of-speech tagging and key phrase extraction, while also using AWS Deep Learning Containers with Amazon SageMaker to build and deploy the custom classifier.

D for me. Question says "The preprocessed text WILL be input to a custom classification algorithm that the data science team has already written and trained using Apache MXNet". So for some reason they want to use MXNet to do the classification, not Amazon Comprehend. So using MXNet for classification is a part of their requirement. How do we meet these requirements quickly? Well, use Amazon Comprehend for part-of-speech and key phrase tasks; and use container for the MXNet stuff.

The question did not make it clear whether the new solution has to use the custom model that the team built or not.

A for me

Agreed with A, Comprehend 3 functions