Model selection for tasks in NLP - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to import the model used for text classification.

NLP

from sklearn.[1] import LogisticRegression

Drag options to blanks, or click blank then click option'

Acluster

Blinear_model

Censemble

Ddecomposition

Attempts:

3 left

💡 Hint

Common Mistakes

Choosing 'ensemble' which contains tree-based models.

Choosing 'cluster' which is for clustering algorithms.

The LogisticRegression model is part of the linear_model module in scikit-learn, commonly used for classification tasks.

2fill in blank

medium

Complete the code to create a model for sentiment analysis using a pretrained transformer.

NLP

from transformers import [1]
model = [1].from_pretrained('distilbert-base-uncased-finetuned-sst-2-english')

Drag options to blanks, or click blank then click option'

ADistilBertForSequenceClassification

BBertModel

CGPT2Model

DRobertaModel

Attempts:

3 left

💡 Hint

Common Mistakes

Using BertModel which is a base model without classification head.

Using GPT2Model which is for text generation.

The DistilBertForSequenceClassification model is designed for sequence classification tasks like sentiment analysis.

3fill in blank

hard

Fix the error in the code to select the right model for named entity recognition (NER).

NLP

from transformers import AutoModelFor[1]
model = AutoModelFor[1].from_pretrained('dbmdz/bert-large-cased-finetuned-conll03-english')

Drag options to blanks, or click blank then click option'

ASequenceClassification

BQuestionAnswering

CTokenClassification

DMaskedLM

Attempts:

3 left

💡 Hint

Common Mistakes

Using SequenceClassification which is for whole sentence labels.

Using MaskedLM which is for language modeling.

Named entity recognition is a token-level task, so the correct model class is AutoModelForTokenClassification.

4fill in blank

hard

Fill both blanks to create a pipeline for text summarization using Hugging Face.

NLP

from transformers import pipeline
summarizer = pipeline(task=[1], model=[2])

Drag options to blanks, or click blank then click option'

A'summarization'

B'translation_en_to_fr'

C'facebook/bart-large-cnn'

D'text-generation'

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'translation_en_to_fr' which is for translation.

Using 'text-generation' which is for generating text, not summarizing.

The task for summarization is 'summarization' and the model 'facebook/bart-large-cnn' is a popular pretrained summarization model.

5fill in blank

hard

Fill all three blanks to build and evaluate a simple text classification model using scikit-learn.

NLP

from sklearn.feature_extraction.text import [1]
from sklearn.linear_model import [2]
from sklearn.metrics import [3]

vectorizer = [1]()
X_train_vec = vectorizer.fit_transform(X_train)
model = [2]()
model.fit(X_train_vec, y_train)
y_pred = model.predict(vectorizer.transform(X_test))
score = [3](y_test, y_pred)

Drag options to blanks, or click blank then click option'

ACountVectorizer

BLogisticRegression

Caccuracy_score

DTfidfVectorizer

Attempts:

3 left

💡 Hint

Common Mistakes

Using TfidfVectorizer instead of CountVectorizer (both work but only one is correct here).

Using other metrics like precision instead of accuracy_score.

CountVectorizer converts text to numbers, LogisticRegression is a simple classifier, and accuracy_score measures how well the model predicts.