Natural Language Processing Cheat Sheet 2026

The 30 highest-yield Natural Language Processing facts, distilled from real exam questions. Print it, save it as a PDF, or study it here — free, no sign-up.

1. What does 'TF-IDF' stand for in NLP? → Term Frequency–Inverse Document Frequency
2. What does text mining stemming mean? → Reducing a word to its root
3. In the ROUGE metric used for text summarization evaluation, what does ROUGE-L specifically measure? → Longest common subsequence between candidate and reference
4. Which evaluation metric computes the geometric mean of precision and recall of n-gram overlaps between a machine translation output and reference translations? → BLEU
5. What is a bag-of-words (BoW) representation? → A text representation that counts word occurrences while ignoring word order
6. What is the purpose of 'masked language modeling' (MLM) as used in BERT's pre-training? → Training the model to predict randomly masked tokens using bidirectional context
7. What is coreference resolution? → Determining when two or more expressions in a text refer to the same real-world entity
8. What is 'transfer learning' in the context of NLP? → Applying knowledge gained from pre-training on large data to specific downstream tasks
9. Which parsing strategy processes input left-to-right and builds the parse bottom-up using a stack and input buffer? → Shift-reduce parsing
10. What is the main advantage of subword tokenization over word-level tokenization? → It handles rare and out-of-vocabulary words by splitting them into known subword units
11. In NLP, what is the 'vocabulary' of a language model? → The fixed set of tokens the model can process and generate
12. What does the 'temperature' parameter control during language model text generation? → The sharpness or randomness of the output token probability distribution
13. What is the role of part-of-speech (POS) tagging in the NLP pipeline? → It assigns grammatical categories (noun, verb, adjective, etc.) to each token
14. What is extractive question answering? → Selecting a span of text from a given passage as the answer to a question
15. What is the primary innovation of the Transformer architecture over RNNs? → It uses self-attention to process all tokens in parallel rather than sequentially
16. What is the main purpose of the CRF (Conditional Random Field) layer commonly added on top of BERT for sequence labeling tasks? → To model dependencies between consecutive output labels
17. Which algorithm is used in the Viterbi decoding process for Hidden Markov Models in NLP? → Dynamic programming
18. What problem does beam search solve in NLP sequence generation tasks? → Suboptimal output due to greedy decoding
19. What is the main limitation of n-gram language models? → They suffer from data sparsity because higher-order n-grams rarely appear in training data
20. Which attention mechanism allows each token in a sequence to attend to all other tokens simultaneously, enabling parallelization? → Self-attention
21. Which of the following is an example of a 'zero-shot' NLP task evaluation? → Asking the model to classify sentiment with no task-specific training examples provided
22. What does 'hallucination' mean when describing large language model outputs? → Producing plausible-sounding but factually incorrect or fabricated information
23. Which loss function is most commonly used for training neural machine translation models? → Cross-entropy loss
24. In dependency parsing, what is the 'head' of a dependency relation? → The word that another word (the dependent) grammatically modifies or depends on
25. Which technique allows a pre-trained language model to adapt to a new task using only a handful of examples shown in the prompt? → Few-shot prompting
26. Morphological segmentation: What is it? → Separate words into individual morphemes and identify the class of the morphemes
27. In the context of question answering, what is a 'reading comprehension' task? → Answering questions by extracting or inferring information from a provided passage
28. What does constituency parsing produce? → A hierarchical tree of phrases grouped by grammatical roles
29. In NLP, what does 'zero-shot learning' refer to? → A model performing a task it was never explicitly trained on
30. What does 'BERT' stand for? → Bidirectional Encoder Representations from Transformers