NLP applications in real world - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Which NLP application is best suited for automatically summarizing long documents?

Imagine you have a very long article and want a short summary that captures the main points. Which NLP application would you use?

A Text Summarization

B Machine Translation

C Named Entity Recognition (NER)

D Sentiment Analysis

💡 Hint

Think about which task reduces text length while keeping important info.

Text Summarization creates a shorter version of a long text, keeping key ideas. NER finds names, Sentiment Analysis finds feelings, and Machine Translation changes language.

❓ Predict Output

intermediate

What is the output of this sentiment analysis code snippet?

Given the code below that uses a simple rule-based sentiment check, what will be the printed output?

NLP

text = "I love sunny days but hate the rain."
positive_words = ['love', 'happy', 'joy']
negative_words = ['hate', 'sad', 'angry']
score = 0
for word in text.lower().split():
    if word in positive_words:
        score += 1
    elif word in negative_words:
        score -= 1
if score > 0:
    print('Positive')
elif score < 0:
    print('Negative')
else:
    print('Neutral')

A Negative

B Neutral

C Positive

D SyntaxError

💡 Hint

Count how many positive and negative words appear exactly in the list.

The text contains 'love' (positive) and 'hate' (negative), so score is 0, resulting in 'Neutral'.

❓ Model Choice

advanced

Which model type is most appropriate for real-time speech-to-text transcription?

You want to convert spoken words into text instantly during a live conversation. Which model type is best suited?

A Recurrent Neural Network (RNN) with attention

B Transformer-based sequence-to-sequence model

C K-Means clustering

D Convolutional Neural Network (CNN) for image classification

💡 Hint

Consider models that handle sequences and context well for language tasks.

Transformer-based seq2seq models are state-of-the-art for speech-to-text, handling long sequences efficiently. RNNs are older and slower, CNNs are for images, and K-Means is unsupervised clustering.

❓ Metrics

advanced

Which metric is most suitable to evaluate a machine translation system?

You built a system that translates English sentences to French. Which metric should you use to measure how good the translations are?

A Accuracy

B Mean Squared Error

C BLEU score

D F1 score

💡 Hint

Think about metrics that compare generated text to reference translations.

BLEU score measures how close the machine translation is to human translations by comparing overlapping words and phrases. Accuracy and F1 are for classification, MSE is for regression.

🔧 Debug

expert

What error does this code raise when running a named entity recognition (NER) pipeline?

Consider this Python code using a popular NLP library to extract named entities. What error will it raise?

NLP

from transformers import pipeline
ner = pipeline('ner')
text = "Apple is looking at buying U.K. startup for $1 billion"
results = ner(text, aggregation_strategy='simple')
print(results)

A TypeError: pipeline() got an unexpected keyword argument 'aggregation_strategy'

B ValueError: aggregation_strategy must be one of ['none', 'simple']

C RuntimeError: Model weights not found

D No error, prints list of entities

💡 Hint

Check if the 'aggregation_strategy' argument is valid for the current transformers version.

In recent versions of transformers, 'aggregation_strategy' is a valid argument for the NER pipeline and will return aggregated entities without error.