NCA-GENL Question 52
Single answerYou are working on a semantic search application where similar documents need to be retrieved based on user queries. To achieve this, you need to generate text embeddings. Which model would be the most appropriate for generating high-quality text embeddings for this use case?
- A
A pre-trained BERT model fine-tuned on a sentence similarity task
- B
A generative GPT model trained on language generation
- C
A convolutional neural network (CNN) designed for image classification
- D
A pre-trained word2vec model trained on a large corpus of raw text
Show answer and explanation
Correct answer: A
Explanation
For generating high-quality text embeddings for a semantic search application, it is critical to use a model fine-tuned for understanding semantic relationships in text. A pre-trained BERT model fine-tuned on a sentence similarity task is ideal because it leverages contextual embeddings to capture nuanced semantic meanings, unlike older models like word2vec or GPT models primarily designed for text generation.
- A. Correct.
A pre-trained BERT model fine-tuned on a sentence similarity task is specifically designed to generate meaningful text embeddings that capture semantic similarity, making it highly suitable for tasks like semantic search.
- B. Incorrect.
A generative GPT model is optimized for text generation rather than directly providing meaningful embeddings for semantic similarity tasks, so it is less suitable for this use case.
- C. Incorrect.
A convolutional neural network (CNN) is designed for image-related tasks, not for processing text or generating text embeddings.
- D. Incorrect.
While a word2vec model can generate word embeddings, it often struggles to capture contextual meaning as effectively as modern transformer-based models like BERT.