Why translation breaks language barriers in NLP - Test Your Understanding

Practice - 5 Tasks

1fill in blank

easy

Complete the code to load a pre-trained translation model from Hugging Face.

NLP

from transformers import MarianMTModel, MarianTokenizer
model_name = 'Helsinki-NLP/opus-mt-en-de'
tokenizer = MarianTokenizer.from_pretrained([1])

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

A'Helsinki-NLP/opus-mt-en-de'

B'bert-base-uncased'

C'facebook/wav2vec2-base-960h'

D'gpt2'

Attempts:

3 left

💡 Hint

Common Mistakes

Using a language model like 'bert-base-uncased' instead of a translation model.

Confusing speech models with translation models.

We use the exact model name 'Helsinki-NLP/opus-mt-en-de' to load the tokenizer for English to German translation.

2fill in blank

medium

Complete the code to tokenize the input text for translation.

NLP

text = 'Hello, how are you?'
tokenized = tokenizer([1], return_tensors='pt', padding=True)

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

Atokenizer

B'Hello, how are you?'

C['Hello', 'how', 'are', 'you']

Dtext

Attempts:

3 left

💡 Hint

Common Mistakes

Passing a list of words instead of a string.

Passing the tokenizer object instead of the text.

We pass the variable 'text' containing the input string to the tokenizer.

3fill in blank

hard

Fix the error in generating the translated tokens.

NLP

translated_tokens = model.generate([1].input_ids, max_length=40)

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

Atext

Btokenized

Cmodel

Dtokenizer

Attempts:

3 left

💡 Hint

Common Mistakes

Passing raw text instead of token IDs.

Passing the tokenizer or model object instead of token IDs.

We use 'tokenized.input_ids' as input to the model's generate function to produce translated tokens.

4fill in blank

hard

Fill both blanks to decode the translated tokens into readable text.

NLP

translated_text = tokenizer.[1](translated_tokens[0], skip_special_tokens=[2])

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

Adecode

BTrue

CFalse

Dencode

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'encode' instead of 'decode'.

Not skipping special tokens, resulting in unwanted symbols.

We use 'decode' to convert tokens back to text and skip_special_tokens=True to remove special tokens.

5fill in blank

hard

Fill all three blanks to create a function that translates English text to German.

NLP

def translate_en_to_de(text):
    inputs = tokenizer(text, return_tensors=[1], padding=True)
    outputs = model.generate(inputs.[2], max_length=50)
    return tokenizer.[3](outputs[0], skip_special_tokens=True)

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

A'pt'

Binput_ids

Cdecode

Dtext

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'text' instead of 'input_ids' for model input.

Returning tokens without decoding.

Using wrong tensor type like 'tf' instead of 'pt'.

We use 'pt' for PyTorch tensors, 'input_ids' for model input, and 'decode' to get the translated text.