Google Professional Machine Learning Engineer Question 134
Select 3Google Cloud PlatformYou are building a text classification model on Vertex AI that processes unstructured text documents from various sources, such as PDFs and Word documents. The data needs to be preprocessed before inference to extract text content. Which of the following steps should you include in your workflow to ensure the text data is ingested correctly into Vertex AI for inference?
- A
Use Google Cloud Document AI to extract structured text from the document files.
- B
Store the raw document files in a Cloud Storage bucket and directly pass the file paths to the Vertex AI model.
- C
Preprocess the extracted text to ensure it matches the input format expected by the Vertex AI model.
- D
Use Cloud Dataflow to handle large-scale preprocessing and transformation of text data if the dataset is large.
- E
Deploy the Vertex AI model first, and then use it to extract text from the documents.
Show answer and explanation
Correct answers: A, C, D
Explanation
To ingest unstructured text data from documents into Vertex AI for inference, you need to extract the text content first using a tool like Google Cloud Document AI. Preprocessing the extracted text is essential to ensure it matches the input format expected by the model. For large datasets, tools like Cloud Dataflow can help manage the preprocessing at scale. Directly using raw document files or expecting Vertex AI to extract text is not feasible.
- A. Correct.
Correct. Google Cloud Document AI is a purpose-built service for extracting structured text from unstructured document files, which is necessary for processing text before ingestion into Vertex AI.
- B. Incorrect.
Incorrect. Passing raw document file paths directly to the Vertex AI model is not supported because Vertex AI models require structured inputs (e.g., text, tabular data, etc.) and cannot process raw files directly.
- C. Correct.
Correct. Preprocessing the extracted text to align with the input format expected by the Vertex AI model is a critical step to ensure the model can process the data correctly.
- D. Correct.
Correct. Cloud Dataflow is an appropriate tool for handling large-scale preprocessing and transformation tasks, especially when working with a sizable dataset.
- E. Incorrect.
Incorrect. Vertex AI models are not designed to extract text from raw documents; this step must be performed before sending inputs to the model.