AIF-C01 Question 62
Single answerA retail company wants to use a transformer-based large language model (LLM) to analyze customer feedback and generate personalized product recommendations. However, they notice that the model sometimes struggles with long inputs, such as detailed customer reviews. What foundational generative AI concept can help them address this issue?
- A
Tokenization
- B
Chunking
- C
Embeddings
- D
Prompt engineering
Show answer and explanation
Correct answer: B
Explanation
Transformer-based LLMs typically have a maximum input token limit, which can cause issues when processing long text inputs like detailed customer reviews. Chunking is the appropriate solution, as it breaks the input into smaller segments that the model can process individually, ensuring no information is lost due to input size constraints.
- A. Incorrect.
Tokenization refers to breaking text into smaller units, such as words or subwords, for processing by the model. While tokenization is important, it does not directly address the issue of handling long inputs.
- B. Correct.
Chunking involves splitting long inputs into smaller, manageable pieces that can be processed by the model. This is a common solution to address limitations in model input size, making it the correct answer.
- C. Incorrect.
Embeddings are vector representations of data that capture semantic meaning. While embeddings are useful in many AI tasks, they are not specifically intended to handle long inputs.
- D. Incorrect.
Prompt engineering involves crafting effective prompts to guide the model's behavior. While it can improve the quality of model outputs, it does not solve the issue of input length limitations.