Fine-grained sentiment (5-class) in NLP - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to load the dataset for fine-grained sentiment analysis.

NLP

from datasets import load_dataset

dataset = load_dataset('sst', '[1]')

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

Atrain

Btest

Call

Dvalidation

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'test' instead of 'train' to load training data.

Using 'all' which is not a valid split.

The 'train' split is used to load the training data for fine-grained sentiment analysis.

2fill in blank

medium

Complete the code to tokenize the input sentences for the model.

NLP

from transformers import AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained('distilbert-base-uncased')
tokens = tokenizer([1], padding=True, truncation=True, return_tensors='pt')

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

Adataset['sentence']

Bdataset['label']

Cdataset['text']

Ddataset['input']

Attempts:

3 left

💡 Hint

Common Mistakes

Trying to tokenize labels instead of sentences.

Using a wrong field name like 'text' or 'input' which may not exist.

The 'sentence' field contains the text to tokenize for sentiment analysis.

3fill in blank

hard

Fix the error in the model output layer to match 5 sentiment classes.

NLP

from transformers import AutoModelForSequenceClassification

model = AutoModelForSequenceClassification.from_pretrained('distilbert-base-uncased', num_labels=[1])

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

A2

B4

C3

D5

Attempts:

3 left

💡 Hint

Common Mistakes

Setting num_labels to 2 or 3 which is for binary or ternary sentiment.

Using 4 which is incorrect for 5-class sentiment.

Fine-grained sentiment analysis has 5 classes, so num_labels must be 5.

4fill in blank

hard

Fill both blanks to compute accuracy and macro F1 score for the 5-class sentiment model.

NLP

from sklearn.metrics import [1], [2]

accuracy = [1](true_labels, predictions)
macro_f1 = [2](true_labels, predictions, average='macro')

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

Aaccuracy_score

Bf1_score

Cprecision_score

Drecall_score

Attempts:

3 left

💡 Hint

Common Mistakes

Using precision_score or recall_score instead of f1_score for F1 metric.

Not specifying average='macro' for multi-class F1.

Use accuracy_score for accuracy and f1_score with macro average for balanced F1 across classes.

5fill in blank

hard

Fill all three blanks to create a dictionary of predictions with labels and filter positive sentiments.

NLP

pred_dict = [1](label: pred for label, pred in zip(labels, predictions) if pred [2] [3])

# Filter only positive sentiments (labels 3 and 4)

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

Adict

B>=

C3

Dlist

Attempts:

3 left

💡 Hint

Common Mistakes

Using list instead of dict for comprehension.

Using wrong comparison operator like '<' or '==' instead of '>='.

Filtering with wrong threshold value.

Use dict comprehension to map labels to predictions and filter predictions >= 3 for positive sentiments.