Đề 6 – Bài tập, đề thi trắc nghiệm online Xử lý ngôn ngữ tự nhiên

1. What is `cross-lingual NLP′ and what are some of its challenges?

A. Cross-lingual NLP is about processing text in very formal or technical language.

B. Cross-lingual NLP deals with NLP tasks that involve multiple languages. Challenges include language variations, lack of parallel data, and the need for language-agnostic representations.

C. Cross-lingual NLP is focused on translating between languages with similar linguistic structures.

D. Cross-lingual NLP is a subset of NLP that only works with European languages.

2. What is `dialogue management′ in the context of building chatbots or conversational agents?

A. Dialogue management is about translating user input into different languages in a chatbot.

B. Dialogue management is the component of a chatbot that decides the next action or response based on the conversation history and current user input, ensuring coherent and contextually relevant interactions.

C. Dialogue management is responsible for the visual design and user interface of a chatbot.

D. Dialogue management is a technique for summarizing long conversations into shorter summaries.

3. Which of the following NLP techniques is most directly concerned with understanding the meaning of words in context?

A. Part-of-speech tagging.

B. Named Entity Recognition.

C. Word Sense Disambiguation.

D. Sentiment Analysis.

4. What is `transfer learning′ in NLP and why is it beneficial?

A. Transfer learning is about transferring text from one format to another (e.g., text to speech).

B. Transfer learning involves using knowledge gained from solving one problem to solve a different but related problem, often improving performance and reducing training time.

C. Transfer learning is the process of translating text between languages.

D. Transfer learning is a method for correcting grammatical errors in text.

5. What is the primary purpose of tokenization in Natural Language Processing (NLP)?

A. To remove punctuation from text.

B. To convert text into numerical vectors.

C. To break down text into smaller units like words or phrases.

D. To identify the language of the text.

6. What is `semantic similarity′ in NLP, and why is it important?

A. Semantic similarity is about how similar the sounds of words are.

B. Semantic similarity measures the degree of meaning relatedness between words, sentences, or documents. It′s important for tasks like information retrieval, text summarization, and question answering to understand meaning beyond surface-level word matching.

C. Semantic similarity is about how similar the length of documents are.

D. Semantic similarity is only relevant for languages with similar vocabulary.

7. What is the purpose of `word embeddings′ like Word2Vec or GloVe in NLP?

A. To compress text data for efficient storage.

B. To represent words as numerical vectors that capture semantic relationships.

C. To generate synonyms for words in a given text.

D. To correct grammatical errors in sentences.

8. What is `zero-shot learning′ in NLP and in what scenarios is it particularly useful?

A. Zero-shot learning is about training models with no data at all.

B. Zero-shot learning refers to the ability of a model to generalize to new tasks or classes without explicit training examples for those specific tasks or classes. It′s useful when labeled data for a task is scarce or unavailable.

C. Zero-shot learning is a method for compressing NLP models to reduce their size.

D. Zero-shot learning is a technique for improving the speed of text processing.

9. What role does `syntax′ play in Natural Language Processing?

A. Syntax focuses on the meaning of words and sentences.

B. Syntax deals with the structure of sentences and the rules governing word order.

C. Syntax is concerned with the sound of language and pronunciation.

D. Syntax refers to the cultural context of language use.

10. Explain the concept of `n-grams′ in NLP and provide an example of their application.

A. N-grams are a type of neural network architecture used in NLP.

B. N-grams are contiguous sequences of n items from a given text. For example, in language modeling, bigrams (2-grams) like `natural language′ can be used to predict the next word in a sequence.

C. N-grams are used for text summarization by selecting the most frequent phrases.

D. N-grams are a method for correcting spelling errors by identifying common word combinations.

11. What is the `bag-of-words′ model in NLP and what is its primary limitation?

A. It′s a model that captures word order effectively, but it′s computationally expensive.

B. It represents text as the collection of its words, disregarding grammar and word order, which is a major limitation.

C. It′s a model for generating text, but it cannot handle long sentences.

D. It′s used for semantic analysis, but it fails to capture sentiment.

12. What is `sentiment analysis′ in NLP and why is it valuable for businesses?

A. Sentiment analysis is about correcting spelling errors; it helps businesses improve communication.

B. Sentiment analysis is about identifying emotions in text; it provides insights into customer opinions and brand perception.

C. Sentiment analysis is about summarizing long documents; businesses use it to save time reading feedback.

D. Sentiment analysis is about translating languages; it helps businesses reach global markets.

13. Which NLP task involves identifying and classifying named entities in text, such as persons, organizations, and locations?

A. Sentiment Analysis.

B. Topic Modeling.

C. Named Entity Recognition (NER).

D. Text Summarization.

14. In NLP, what does `disambiguation′ mean, and why is it a crucial step in many applications?

A. Disambiguation is about making text more ambiguous to protect privacy.

B. Disambiguation is the process of resolving ambiguity in language, such as word sense disambiguation or resolving pronoun references. It′s crucial because ambiguity can lead to misinterpretations and errors in NLP applications.

C. Disambiguation is about simplifying complex sentences into simpler forms.

D. Disambiguation is a technique for removing irrelevant information from text.

15. What is the `semantic gap′ in NLP and how do techniques like word embeddings attempt to address it?

A. The semantic gap refers to the difference in processing speed between different NLP models.

B. The semantic gap is the discrepancy between the symbolic representation of language and its underlying meaning. Word embeddings bridge this gap by providing distributed representations that capture semantic relationships.

C. The semantic gap is the difficulty in translating between languages with very different grammars.

D. The semantic gap is the problem of identifying named entities in noisy text data.

16. What is the role of `stop words′ in NLP and how are they typically handled?

A. Stop words are crucial for understanding sentence structure and are always kept.

B. Stop words are high-frequency words that are often removed to reduce noise and improve processing efficiency.

C. Stop words are words that indicate sentiment and are essential for sentiment analysis.

D. Stop words are only relevant in languages with complex grammar.

17. What are `Transformer networks′ in NLP and what advantages do they offer over Recurrent Neural Networks (RNNs) for sequence processing?

A. Transformer networks are simpler to train but less accurate than RNNs.

B. Transformer networks use attention mechanisms and can process sequences in parallel, overcoming the sequential processing limitation of RNNs and enabling better handling of long-range dependencies.

C. Transformer networks are only suitable for text classification tasks, unlike RNNs which are more versatile.

D. Transformer networks require significantly more training data than RNNs to achieve comparable performance.

18. Which of the following is an example of a `sequence-to-sequence′ task in NLP?

A. Sentiment classification of movie reviews.

B. Spam detection in emails.

C. Machine Translation.

D. Topic extraction from news articles.

19. What is `active learning′ in NLP and why might it be preferred over passive learning in certain situations?

A. Active learning is about learning actively by reading books and articles, unlike passive learning which is just listening to lectures.

B. Active learning is a machine learning approach where the model strategically selects the most informative data points to be labeled by a human annotator, which can lead to better performance with less labeled data compared to passive learning (random sampling).

C. Active learning is a technique for making language models more interactive and conversational.

D. Active learning refers to models that can learn continuously without forgetting previous knowledge.

20. What is the role of `attention mechanisms′ in modern NLP models, particularly in transformers?

A. Attention mechanisms are used to reduce the dimensionality of word embeddings.

B. Attention mechanisms allow the model to focus on relevant parts of the input sequence when processing or generating output, improving performance on tasks like translation and summarization.

C. Attention mechanisms are used to correct grammatical errors in the input text.

D. Attention mechanisms are responsible for tokenizing the input text.

21. What is `Text Summarization′ in NLP, and what are the two main approaches to it?

A. Text summarization is about translating text into a shorter format in the same language. The two main approaches are translation and paraphrasing.

B. Text summarization is the process of creating a concise and coherent summary of a longer text. The two main approaches are extractive summarization (selecting existing sentences) and abstractive summarization (generating new sentences).

C. Text summarization is about correcting grammatical errors and improving text quality. The two approaches are rule-based and statistical methods.

D. Text summarization is about identifying the main topics in a document. The two approaches are topic modeling and keyword extraction.

22. What is `knowledge graph′ in the context of NLP and how is it used?

A. A knowledge graph is a visual representation of text sentiment.

B. A knowledge graph is a structured representation of facts and entities and their relationships, extracted from text or other sources. It′s used for tasks like question answering, information retrieval, and reasoning.

C. A knowledge graph is a type of word embedding technique.

D. A knowledge graph is used for translating text between multiple languages simultaneously.

23. Stemming and lemmatization are both used to reduce words to their base form. What is the key distinction between these two techniques?

A. Stemming uses a dictionary, while lemmatization uses rules.

B. Lemmatization always produces a valid word as the base form, while stemming might not.

C. Stemming is more computationally intensive than lemmatization.

D. Lemmatization is applied before tokenization, while stemming is applied after.

24. What is the main challenge that `Part-of-Speech (POS) tagging′ aims to solve in NLP?

A. Identifying the topic of a document.

B. Determining the sentiment expressed in a sentence.

C. Assigning grammatical tags (like noun, verb, adjective) to each word in a sentence.

D. Translating text into another language.

25. What is `language modeling′ in NLP and what is its primary goal?

A. Language modeling is about translating text into different languages.

B. Language modeling is the task of predicting the probability of a sequence of words. Its primary goal is to learn the patterns and structure of a language to generate or evaluate text.

C. Language modeling is about summarizing long documents into shorter versions.

D. Language modeling is a technique for correcting grammatical errors in text.

26. What is `topic modeling′ in NLP and what kind of insights can it provide?

A. Topic modeling is used to translate documents into different languages.

B. Topic modeling is a technique to discover abstract `topics′ that occur in a collection of documents, revealing thematic patterns.

C. Topic modeling is used to correct grammatical errors in text.

D. Topic modeling is a method for summarizing long documents into shorter versions.

27. In the context of machine translation, what does `BLEU′ score measure?

A. The grammatical correctness of the translated text.

B. The fluency and readability of the translated text.

C. The similarity between the machine-translated text and human reference translations.

D. The computational efficiency of the translation model.

28. What is the `Curse of Dimensionality′ in the context of NLP and machine learning?

A. It refers to the difficulty of processing very long sentences.

B. It describes the problem of having too few data points for the number of features, leading to sparse data and model overfitting.

C. It′s the challenge of dealing with noisy or ambiguous data in natural language.

D. It′s the computational cost of training very large language models.

29. What are some of the ethical concerns associated with the use of large language models like GPT-3?

A. Primarily concerns about the computational cost and energy consumption.

B. Ethical concerns include potential for bias amplification, generation of misinformation, misuse for malicious purposes, and job displacement in certain language-related professions.

C. The main ethical concern is the risk of these models becoming sentient.

D. Ethical concerns are minimal as these models are just tools and depend on human usage.

30. Explain the concept of `regular expressions′ in NLP and their typical use cases.

A. Regular expressions are a type of neural network used for sequence processing.

B. Regular expressions are patterns used to match character combinations in text. They are used for tasks like text cleaning, information extraction (e.g., finding email addresses), and tokenization based on patterns.

C. Regular expressions are used for sentiment analysis to detect emotional words.

D. Regular expressions are a method for translating text from one language to another based on predefined rules.