Part-of-speech tagging in NLP - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

What is the main goal of part-of-speech tagging?

Imagine you have a sentence and you want to label each word with its role, like noun or verb. What is the main goal of part-of-speech tagging in this context?

A To count the number of words in the sentence.

B To translate the sentence into another language word by word.

C To assign each word in a sentence its grammatical category such as noun, verb, adjective, etc.

D To remove all punctuation marks from the sentence.

💡 Hint

Think about labeling words with their roles in a sentence.

Part-of-speech tagging assigns grammatical categories to each word, helping understand sentence structure.

❓ Predict Output

intermediate

Output of POS tagging with NLTK

What is the output of this Python code using NLTK for POS tagging?

NLP

import nltk
nltk.download('averaged_perceptron_tagger')
sentence = 'The quick brown fox jumps over the lazy dog'.split()
tagged = nltk.pos_tag(sentence)
print(tagged)

A [('The', 'DT'), ('quick', 'NN'), ('brown', 'VB'), ('fox', 'JJ'), ('jumps', 'NN'), ('over', 'IN'), ('the', 'DT'), ('lazy', 'NN'), ('dog', 'VB')]

B [('The', 'DT'), ('quick', 'JJ'), ('brown', 'JJ'), ('fox', 'NN'), ('jumps', 'VBZ'), ('over', 'IN'), ('the', 'DT'), ('lazy', 'JJ'), ('dog', 'NN')]

C [('The', 'PRP'), ('quick', 'RB'), ('brown', 'NN'), ('fox', 'NN'), ('jumps', 'NN'), ('over', 'IN'), ('the', 'DT'), ('lazy', 'JJ'), ('dog', 'VB')]

D [('The', 'NN'), ('quick', 'VB'), ('brown', 'NN'), ('fox', 'VB'), ('jumps', 'NN'), ('over', 'VB'), ('the', 'NN'), ('lazy', 'VB'), ('dog', 'NN')]

💡 Hint

Look for common POS tags: DT (determiner), JJ (adjective), NN (noun), VBZ (verb, 3rd person singular present), IN (preposition).

The NLTK POS tagger assigns tags like DT for 'The', JJ for adjectives like 'quick', NN for nouns like 'fox', and VBZ for verbs like 'jumps'.

❓ Model Choice

advanced

Best model type for accurate POS tagging on new text

You want to build a part-of-speech tagger that works well on new sentences it has never seen before. Which model type is generally best for this task?

A A neural network model trained on a large annotated corpus with word embeddings

B A simple frequency-based tagger that assigns the most common tag per word

C A rule-based system with manually written grammar rules

D A random tagger that assigns tags randomly to each word

💡 Hint

Think about models that learn patterns from lots of examples and generalize well.

Neural network models with word embeddings capture complex patterns and context, making them best for accurate POS tagging on new text.

❓ Metrics

advanced

Which metric best measures POS tagging accuracy?

After training a POS tagger, you want to evaluate how well it labels words correctly. Which metric is most appropriate?

A Precision of predicted tags over all words

B F1 score of predicted tags ignoring true negatives

C Recall of predicted tags over all words

D Overall accuracy: percentage of words tagged correctly

💡 Hint

Think about how many words got the right tag out of all words.

Overall accuracy measures the percentage of words correctly tagged, which is the standard metric for POS tagging.

🔧 Debug

expert

Why does this POS tagging code raise an error?

Consider this code snippet using spaCy for POS tagging. Why does it raise an AttributeError?

NLP

import spacy
nlp = spacy.load('en_core_web_sm')
doc = nlp('I love machine learning')
for token in doc:
    print(token.pos)

A Because <code>token.pos</code> is a property but returns an integer code, not a string tag

B Because <code>token.pos</code> is a method and needs parentheses to be called

C Because the model <code>en_core_web_sm</code> is not installed

D Because <code>token.pos</code> does not exist; the correct attribute is <code>token.tag</code>

💡 Hint

Check the spaCy documentation for token.pos and what it returns.

token.pos returns an integer enum code, not a string. To get the string tag, use token.pos_ (with underscore).