Question Bank
NLP interview questions
Text classification and document retrieval call for different choices. Use these questions to review text representations, language-model training and evaluation. Distinguish properties of the model from the data and the way it is used.
Showing 20 of 76 questions
Compare the tokenization methods BPE, WordPiece, and SentencePiece. What are their main differences?
Why is positional encoding important in transformers?
Explain how the self-attention mechanism works in transformers and how it differs from cross-attention.
What are the advantages and disadvantages of using GPT compared to BERT?
When should the RAG (Retrieval-Augmented Generation) approach be used in text processing tasks?
How does Reinforcement Learning from Human Feedback (RLHF) work and where can it be useful?
Explain how the context-dependent word vector representation method Word2Vec works.
What is BERT and how do its Masked Language Modeling (MLM) and Next Sentence Prediction (NSP) mechanisms affect its performance?
Compare the tokenization approaches of BPE and WordPiece. What are their main differences?
How does the attention mechanism work in transformers, and what is the difference between self-attention and cross-attention?
When and why should one use the BM25 method over TF-IDF in search tasks?
What is FastText and what advantages does it offer compared to other word embedding models?
How do text generation methods like temperature sampling and top-k sampling work? When should each be used?
What is the role of positional encoding in transformers, and what are its variants?
Explain how ELMo works and how it differs from other word embedding models.
Compare BM25 and TF-IDF. In which cases is it better to use each of them?
Prepare for your next interview with Vibe Interview.
Download Vibe Interview