NLP-2

What is an N-gram in word-level analysis?

• A method for visualizing data

• A contiguous sequence of N words from a text

• A machine learning model for classification

• A type of encryption algorithm

In the context of N-grams, what does an "unsmoothed N-gram" model lack?

• Accurate representation of word frequencies

• A mechanism to handle zero probabilities for unseen word combinations

• The ability to create sequences of words

• The ability to identify unique words

What is a major limitation of unsmoothed N-grams?

• They are computationally efficient.

• They fail to account for unseen word combinations, assigning zero probability.

• They require preprocessing of data.

• They always generate incorrect probabilities for frequent words.

What is the purpose of smoothing in N-gram models?

• To improve the visualization of text

• To assign non-zero probabilities to unseen word combinations

• To identify the most frequent words in a text

• To reduce the length of the N-grams

Which of the following is a commonly used smoothing technique?

• Bagging

• Laplace (Add-One) Smoothing

• Principal Component Analysis (PCA)

• Tokenization

How does Laplace (Add-One) Smoothing work?

• It adds a constant to all probabilities to normalize them.

• It adds one to the count of each word or N-gram and recalculates probabilities.

• It removes all low-frequency words from the analysis.

• It reduces the size of the vocabulary.

Which smoothing technique is more advanced and accounts for the probability of unseen events in N-grams?

• Laplace Smoothing

• Backoff Smoothing

• N-gram Filtering

• Token Replacement

What does "backoff" refer to in Backoff Smoothing?

• Reducing the size of N-grams to single words

• Assigning lower probabilities to frequent words

• Falling back to smaller N-grams (e.g., bigrams or unigrams) when a higher-order N-gram is unseen

• Ignoring the context of words entirely

Which smoothing technique uses interpolation between different N-gram levels?

• Laplace Smoothing

• Backoff Smoothing

• Interpolated Smoothing

• Zero Smoothing

What is the primary purpose of interpolation in language modeling?

• To preprocess text data before analysis

• To combine probabilities from different N-gram levels for better accuracy

• To identify word classes in a corpus

• To normalize probabilities in text