PMLE Practice Test 17 — 15 Questions

Question 1

A company wants to transcribe audio from customer service calls and then analyze the sentiment of the transcribed text. Which TWO Google Cloud services should they use?

Accepted Answer

Natural Language API. Speech-to-Text transcribes audio to text, and Natural Language API can analyze sentiment from text. Vision API is for images, Translation for language translation, and Document AI for document processing.

Answer

Document AI

Answer

Translation API

Answer

Vision API

Question 2

A retail company wants to implement a recommendation system using Recommendations AI. They need to generate personalized recommendations for users based on their browsing history and purchase behavior. Which THREE recommendation types are available in Recommendations AI?

Accepted Answer

recommended-for-you. Recommendations AI offers five recommendation types: recommended-for-you (personalized), others-you-may-like (item similarity), frequently-bought-together (complementary), recently-viewed (session-based), and also-viewed. 'Most popular' is not a built-in type.

Answer

trending-now

Answer

most-popular

Question 3

A data engineer is using BigQuery ML with a BOOSTED_TREE_CLASSIFIER model. After training, they want to evaluate the model and understand which features contribute most to predictions. Which THREE BigQuery ML functions should they use?

Accepted Answer

ML.EVALUATE. ML.EVALUATE evaluates model performance, ML.FEATURE_IMPORTANCE returns feature importance scores, and ML.PREDICT makes predictions. ML.TRAIN is for training, ML.GLOBAL_EXPLAIN is not a standard function (ML.EXPLAIN_PREDICT exists but is not listed as an option).

Answer

ML.GLOBAL_EXPLAIN

Answer

ML.TRAIN

Question 4

A team trained a TensorFlow model locally and wants to deploy it to BigQuery ML for predictions without retraining. They have exported the SavedModel to Cloud Storage. Which statement is correct?

Accepted Answer

They can create a model using CREATE MODEL with model_type='tensorflow' and the path to the SavedModel.. BigQuery ML supports importing TensorFlow SavedModels with model_type='tensorflow'. ML.TRAIN is for training, not importing. ML.PREDICT requires a model in BigQuery, not directly on Cloud Storage. No conversion is needed.

Answer

They need to convert the model to a BigQuery ML native format first.

Answer

They must first retrain the model using ML.TRAIN on BigQuery.

Answer

They can use ML.PREDICT directly on the SavedModel in Cloud Storage.

Question 5

A retail company uses Recommendations AI to power personalized product recommendations on their website. They notice that the 'frequently-bought-together' model is not capturing complementary items that are often purchased in the same session but not necessarily in the same transaction. Which TWO actions should they take to improve the model?

Accepted Answer

Enable the 'others-you-may-like' recommendation type in addition to 'frequently-bought-together'. To capture cross-session patterns, the company should ensure that user events (including session-level co-occurrence) are properly tracked and ingested. Enabling session-level events and using the 'others-you-may-like' model (which uses co-viewed behavior) can help. Setting the recommendation type to 'frequently-bought-together' does not address session-level data. Reducing event retention would harm model quality.

Answer

Decrease the event retention period to focus on recent purchases

Answer

Use AutoML Tables to build a custom recommendation model

Answer

Set the recommendation type to only 'frequently-bought-together'

Question 6

A data analyst wants to train a binary classification model in BigQuery ML on a dataset of 10 million rows with 50 features. They need to evaluate the model's performance on a held-out test set. Which sequence of SQL statements should they run?

Accepted Answer

CREATE MODEL then ML.EVALUATE. First, create the model using CREATE MODEL. Then, evaluate it using ML.EVALUATE, which uses the test split defined in the model options. ML.PREDICT is for predictions, not evaluation. ML.TRAIN is not a valid function; model training is done via CREATE MODEL.

Answer

CREATE MODEL then ML.FEATURE_IMPORTANCE

Answer

ML.TRAIN then ML.EVALUATE

Answer

CREATE MODEL then ML.PREDICT

Question 7

A media company wants to automatically moderate user-uploaded videos by detecting explicit content (e.g., violence, adult material). They need a solution that integrates with their video processing pipeline and scales to millions of videos. Which approach should they take?

Accepted Answer

Use Video Intelligence API with explicit content detection. Video Intelligence API provides explicit content detection as a pre-built feature. It can analyze video content and flag inappropriate material. AutoML Video would require custom training, which is unnecessary. Vision API is for images. Natural Language API is for text.

Answer

Use AutoML Video to train a custom explicit content detection model

Answer

Use Natural Language API on video transcripts

Answer

Use Vision API to analyze each video frame

Question 8

A company has a TensorFlow model trained outside of Google Cloud and wants to use it for online predictions on Vertex AI. They have saved the model in SavedModel format. What is the most efficient way to deploy this model?

Accepted Answer

Upload the saved model to Vertex AI and create an endpoint for online predictions. Vertex AI supports importing SavedModel directly without retraining. BigQuery ML can import TensorFlow models but is for batch predictions. Vertex AI Prediction is the standard for online predictions. AutoML Tables is for training new models.

Answer

Import the model into BigQuery ML using CREATE MODEL with model_type='TENSORFLOW'

Answer

Use Vertex AI AutoML Tables to retrain the model

Answer

Use Cloud Functions to run the model for each prediction request

Question 9

A financial institution needs to extract structured data from scanned PDFs of loan applications, including text fields and tables. They require a human review step for high-risk applications. Which Google Cloud service and configuration should they use?

Accepted Answer

Document AI with a form parser processor and enable Human-in-the-Loop for high-risk applications. Document AI provides specialised processors for document parsing (including form parser). Human-in-the-Loop (HITL) is a feature of Document AI that allows human review for high-risk documents. Cloud Vision API is for image analysis, not document parsing. Vertex AI AutoML Vision is for image classification/object detection.

Answer

Document AI with a custom extractor processor and use Cloud Functions for human review

Answer

Cloud Vision API to detect text and tables, then send to Cloud Dataflow for processing

Answer

Vertex AI AutoML Vision to train a custom model for document parsing

Question 10

A company is building a document processing pipeline using Document AI to extract data from invoices. They want to ensure high accuracy and handle edge cases where the model may be uncertain. Which THREE steps should they include in their pipeline?

Accepted Answer

Regularly retrain the processor using human-verified data. To improve accuracy, use human review for low-confidence documents, create a custom processor (or use a pre-built one) tailored to their invoice format, and continuously improve the model with human-labeled data. Using the form parser is a good start but not enough; they need to enable HITL and potentially train a custom processor.

Answer

Use the pre-built invoice parser without any modifications

Answer

Use AutoML Vision to classify invoice types

Question 11

A company wants to transcribe customer service calls in real-time. The audio is telephony quality (8 kHz). Which Speech-to-Text model should they specify?

Accepted Answer

telephony. The 'telephony' model is optimised for audio from phone calls (8 kHz). 'latest_long' is for long-form non-telephony audio, 'latest_short' for short utterances, and there is no 'default' model.

Answer

latest_short

Answer

latest_long

Answer

default

Question 12

A company needs to analyze customer feedback from app reviews to identify common themes and sentiment. They have millions of reviews in multiple languages. Which combination of pre-built APIs should they use?

Accepted Answer

Translation API then Natural Language API. Natural Language API can perform sentiment analysis and entity extraction. Translation API can translate non-English reviews to English before analysis. Speech-to-Text is for audio. Vision API is for images.

Answer

Speech-to-Text then Natural Language API

Answer

Natural Language API only

Answer

Vision API then Natural Language API

Question 13

You are using DVC for data versioning in an ML project on Google Cloud. Your training data is stored in Cloud Storage. You want to track a new version of the dataset after preprocessing. Which DVC command should you use to register the changes?

Accepted Answer

dvc add data/processed. DVC tracks data versions via 'dvc add' which creates a .dvc file that points to the data in Cloud Storage. 'dvc run' is for pipelines, 'dvc push' uploads cached data to remote storage, and 'dvc commit' saves changes to DVC-tracked files after a pipeline run.

Answer

dvc push

Answer

dvc run -n preprocess

Answer

dvc commit

Question 14

You need to create a reproducible snapshot of a BigQuery table as of a specific timestamp for ML model training. The snapshot should be queryable without copying the entire dataset. Which BigQuery feature should you use?

Accepted Answer

BigQuery table snapshots. BigQuery table snapshots create a read-only copy of a table at a specific point in time, queryable without additional storage costs for the data. Time travel queries access historical data but are limited to 7 days, and copying to a new table duplicates storage.

Answer

BigQuery export to Cloud Storage as Parquet

Answer

BigQuery time travel (FOR SYSTEM_TIME AS OF)

Answer

CREATE TABLE AS SELECT with WHERE clause

Question 15

You are setting up feature monitoring in Vertex AI Feature Store to detect drift in a numerical feature. The monitoring job should run daily and alert if the Jensen-Shannon divergence exceeds 0.1. Which configuration should you use?

Accepted Answer

Configure feature monitoring in the feature view with a drift threshold of 0.1 using Jensen-Shannon divergence. Feature monitoring in Vertex AI Feature Store uses monitoring_config with drift detection via statistical tests like Jensen-Shannon divergence. The correct approach is to set the drift threshold in the feature view's monitoring configuration.

Answer

Use BigQuery scheduled queries to compare distributions and send alerts

Answer

Set up a Cloud Composer DAG to compute drift and publish to Cloud Monitoring

Answer

Enable model monitoring on the Vertex AI endpoint to detect drift