NCA-GENM Question 92
Select 3You are tasked with building a multimodal generative AI model that combines text data and image data. The text data is in multiple languages, and the image data has inconsistent dimensions. Which of the following steps are essential to manage and preprocess these data sources before training the model?
- A
Normalize image dimensions to a consistent size.
- B
Translate all text data into a single language.
- C
Randomly shuffle the dataset without any preprocessing.
- D
Tokenize the text data into numerical representations.
- E
Resize the images but leave text data as is, since text doesn't require preprocessing.
Show answer and explanation
Correct answers: A, B, D
Explanation
To preprocess multimodal data effectively, both text and image data must be standardized and converted into formats that are compatible with machine learning models. For images, normalizing dimensions ensures uniformity, while for text, translation and tokenization standardize and prepare the data for processing. Random shuffling or ignoring text preprocessing is insufficient to address the inherent challenges of multimodal datasets.
- A. Correct.
Correct. Normalizing image dimensions ensures that the input image data is consistent for the model, which is crucial for efficient training and performance.
- B. Correct.
Correct. Translating text data into a single language standardizes the linguistic features, reducing complexity and ensuring consistency.
- C. Incorrect.
Incorrect. Randomly shuffling the dataset is useful for training, but it does not address preprocessing issues like data consistency or feature extraction.
- D. Correct.
Correct. Tokenizing text data converts it into a numerical format that can be processed by machine learning models.
- E. Incorrect.
Incorrect. While resizing images is essential, leaving text data unprocessed (e.g., untranslated or untokenized) would prevent the model from effectively learning text features.