Named entity recognition in NLP - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

What is the main goal of Named Entity Recognition (NER)?

Choose the best description of what Named Entity Recognition does in text processing.

A Translate text from one language to another automatically.

B Summarize long documents into short paragraphs.

C Identify and classify key information like names, places, and dates in text.

D Detect the sentiment or emotion expressed in text.

💡 Hint

Think about what entities like people or locations are in a sentence.

NER finds specific words or phrases that represent real-world objects such as people, organizations, or locations and labels them accordingly.

❓ Predict Output

intermediate

Output of NER model prediction on a sentence

Given the following code using spaCy to perform NER, what is the printed output?

NLP

import spacy
nlp = spacy.load('en_core_web_sm')
doc = nlp('Apple is looking at buying U.K. startup for $1 billion')
for ent in doc.ents:
    print(ent.text, ent.label_)

A

Apple ORG
U.K. GPE
$1 billion MONEY

B

Apple ORG
U.K. LOC
$1 billion MONEY

C

Apple PERSON
U.K. LOC
$1 billion QUANTITY

D

Apple ORG
U.K. ORG
$1 billion MONEY

💡 Hint

ORG means organization, GPE means geopolitical entity, MONEY means monetary values.

spaCy's small English model labels 'Apple' as an organization (ORG), 'U.K.' as a geopolitical entity (GPE), and '$1 billion' as money (MONEY).

❓ Model Choice

advanced

Choosing the best model architecture for NER

Which model architecture is most suitable for Named Entity Recognition tasks?

A K-Means clustering for unsupervised grouping

B Convolutional Neural Network (CNN) for image classification

C Generative Adversarial Network (GAN) for data generation

D Recurrent Neural Network (RNN) or Transformer-based models with token-level classification

💡 Hint

NER requires understanding sequences of words and their context.

NER models typically use RNNs or Transformers to process sequences and classify each token into entity categories.

❓ Metrics

advanced

Evaluating NER model performance

Which metric is most appropriate to evaluate the quality of a Named Entity Recognition model?

A Precision, Recall, and F1-score on entity-level

B BLEU score

C Accuracy on sentence classification

D Mean Squared Error (MSE)

💡 Hint

NER evaluation focuses on correctly identifying and classifying entities.

Precision, Recall, and F1-score measure how well the model finds and labels entities correctly, which is essential for NER.

🔧 Debug

expert

Why does this NER model fail to recognize entities correctly?

Consider this Python code snippet using a custom NER model. After training, it predicts no entities on test sentences. What is the most likely cause?

NLP

from transformers import AutoTokenizer, AutoModelForTokenClassification, Trainer, TrainingArguments
from datasets import load_dataset

model_name = 'bert-base-cased'
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForTokenClassification.from_pretrained(model_name, num_labels=9)

# Dataset loading and tokenization omitted for brevity

training_args = TrainingArguments(output_dir='./results', num_train_epochs=3)
trainer = Trainer(model=model, args=training_args, train_dataset=train_dataset, eval_dataset=eval_dataset)
trainer.train()

# Prediction on test sentence
inputs = tokenizer('Microsoft was founded by Bill Gates.', return_tensors='pt')
outputs = model(**inputs)
predictions = outputs.logits.argmax(dim=-1)
print(predictions)

A The tokenizer is incompatible with the model architecture causing prediction errors.

B The model was not fine-tuned on labeled NER data, so it cannot predict entities correctly.

C The number of labels is set incorrectly to 9 instead of 2.

D The input sentence is too short for the model to detect entities.

💡 Hint

Pretrained models need fine-tuning on task-specific data to perform well.

Without fine-tuning on labeled NER data, the model's predictions are random or default, resulting in no meaningful entity recognition.