Multilingual models in NLP - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Why use multilingual models?

Which of the following is the main advantage of using a multilingual model over separate monolingual models?

A It eliminates the need for any language-specific preprocessing steps.

B It always achieves higher accuracy on every language than monolingual models.

C It requires less total training data because it shares knowledge across languages.

D It guarantees zero bias in all languages it supports.

💡 Hint

Think about how sharing information between languages can help when some languages have less data.

Multilingual models share parameters across languages, allowing them to learn from more data collectively. This helps especially low-resource languages. However, they do not always outperform monolingual models on every language, nor do they remove all preprocessing or bias.

❓ Predict Output

intermediate

Output of tokenization in a multilingual model

Given the following code snippet using a multilingual tokenizer, what is the output tokens list?

NLP

from transformers import AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained('bert-base-multilingual-cased')
text = 'Hello, 你好, مرحبا'
tokens = tokenizer.tokenize(text)
print(tokens)

A ['Hello', ',', '你', '好', ',', 'م', '##رح', '##با']

B ['Hello', ',', '你好', ',', 'مرحبا']

C ['Hello', ',', '你', '##好', ',', 'م', '##رح', '##با']

D ['Hello', ',', '你', '好', ',', 'مرحبا']

💡 Hint

Remember that BERT uses WordPiece tokenization which may split words into subwords marked by '##'.

The multilingual BERT tokenizer splits Chinese characters individually and Arabic words into subwords with '##'. So '你' and '好' are separate tokens, and 'مرحبا' is split into 'م', '##رح', '##با'.

❓ Model Choice

advanced

Choosing a model for low-resource languages

You want to build a text classification model for a low-resource language with very little labeled data. Which model choice is best?

A Train a monolingual model from scratch only on the low-resource language data.

B Use a large multilingual pretrained model fine-tuned on the low-resource language data.

C Use a monolingual pretrained model from a high-resource language and translate data.

D Train a multilingual model from scratch on all available languages.

💡 Hint

Consider how pretrained knowledge from many languages can help when labeled data is scarce.

Large multilingual pretrained models have learned shared representations across languages and can transfer knowledge to low-resource languages during fine-tuning, improving performance. Training from scratch or translating data may be less effective.

❓ Metrics

advanced

Evaluating multilingual model performance

You evaluate a multilingual model on three languages and get these F1 scores: English 0.85, Spanish 0.80, Swahili 0.60. Which metric best summarizes overall performance fairly?

A Weighted-average F1 score (weighted by number of samples per language).

B Micro-average F1 score (global average considering all samples).

C Macro-average F1 score (average of all languages' F1 scores).

D Maximum F1 score among the three languages.

💡 Hint

Think about how to fairly combine scores when languages have different data sizes.

Weighted-average F1 accounts for the number of samples per language, giving a fair overall performance measure when data sizes differ. Macro-average treats all languages equally regardless of size, micro-average pools all samples ignoring language, and max ignores other languages.

🔧 Debug

expert

Debugging poor performance on a multilingual model

You fine-tuned a multilingual model on a dataset with English and French texts. The model performs well on English but poorly on French. Which is the most likely cause?

A The optimizer used is incompatible with multilingual models.

B The model architecture does not support French language tokens.

C The training data for French was accidentally excluded during fine-tuning.

D The tokenizer was not properly set to the multilingual tokenizer, causing French text to be tokenized incorrectly.

💡 Hint

Check if the tokenizer matches the model and supports all languages in your data.

If the tokenizer is not set to a multilingual tokenizer, French text may be split incorrectly, leading to poor model input and performance. Model architectures generally support all tokens from their tokenizer. Data exclusion or optimizer incompatibility are less common causes.

Practice

1. What is the main advantage of using a multilingual model in natural language processing?

easy

A. It can understand and process multiple languages with a single model.

B. It requires training a separate model for each language.

C. It only works for English language tasks.

D. It uses more resources than training individual models.

Solution

1. Step 1: Understand the purpose of multilingual models

   Multilingual models are designed to handle many languages using one model instead of separate ones.

2. Step 2: Compare advantages

   This approach saves time and resources by avoiding multiple models for different languages.

3. Final Answer:

   It can understand and process multiple languages with a single model. -> Option A

4. Quick Check:

   Multilingual model advantage = single model for many languages [OK]

Hint: Multilingual means one model for many languages [OK]

Common Mistakes:

• Thinking multilingual models only work for English
• Assuming separate models are needed per language
• Believing multilingual models use more resources

2. Which of the following is the correct way to load a multilingual model using Hugging Face Transformers in Python?

easy

A. model = AutoModel.from_pretrained('xlm-roberta-base')

B. model = AutoModel.from_pretrained('gpt2')

C. model = AutoModel.from_pretrained('bert-base-uncased')

D. model = AutoModel.from_pretrained('bert-large-cased')

Solution

1. Step 1: Identify multilingual model names

   'xlm-roberta-base' is a well-known multilingual model supporting many languages.

2. Step 2: Check other options

   'bert-base-uncased' and 'bert-large-cased' are English-only models; 'gpt2' is a generative English model.

3. Final Answer:

   model = AutoModel.from_pretrained('xlm-roberta-base') -> Option A

4. Quick Check:

   Multilingual model name = 'xlm-roberta-base' [OK]

Hint: Look for 'xlm' or 'multilingual' in model name [OK]

Common Mistakes:

• Choosing English-only models for multilingual tasks
• Confusing generative models with multilingual encoders
• Using model names without checking language support

3. Consider this Python code using Hugging Face Transformers:

from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained('xlm-roberta-base')
model = AutoModelForSequenceClassification.from_pretrained('xlm-roberta-base')
inputs = tokenizer('Bonjour, comment ça va?', return_tensors='pt')
outputs = model(**inputs)
print(outputs.logits.shape)

What will be the printed output shape?

medium

A. torch.Size([1, 1])

B. torch.Size([1, 2])

C. torch.Size([1, 512])

D. torch.Size([1, 768])

Solution

1. Step 1: Understand model type and output

   The model is for sequence classification, which outputs logits for each class. The default 'xlm-roberta-base' classification head has 2 classes.

2. Step 2: Determine output shape

   Batch size is 1 (one sentence), so output logits shape is [1, 2].

3. Final Answer:

   torch.Size([1, 2]) -> Option B

4. Quick Check:

   Sequence classification logits shape = [batch, classes] = [1, 2] [OK]

Hint: Classification logits shape = batch size x number of classes [OK]

Common Mistakes:

• Confusing hidden size with output logits shape
• Assuming output shape matches input token length
• Ignoring batch size dimension

4. You tried to use a multilingual model but got this error:
ValueError: Tokenizer does not have a pad token.
What is the best way to fix this error?

medium

A. Use a different model that does not require padding.

B. Add padding=True when calling the tokenizer.

C. Manually set the pad token with tokenizer.pad_token = tokenizer.eos_token.

D. Ignore the error and continue training.

Solution

1. Step 1: Understand the error cause

   The tokenizer lacks a pad token, which is needed to pad sequences to the same length.

2. Step 2: Fix by assigning pad token

   Assigning the pad token to an existing token like eos_token solves the issue.

3. Final Answer:

   Manually set the pad token with tokenizer.pad_token = tokenizer.eos_token. -> Option C

4. Quick Check:

   Set pad token manually to fix padding error [OK]

Hint: Set pad token manually if missing in tokenizer [OK]

Common Mistakes:

• Ignoring padding requirement
• Trying to skip padding without fixing tokenizer
• Switching models unnecessarily

5. You want to build a multilingual sentiment analysis system supporting English, Spanish, and French. Which approach best balances accuracy and resource use?

hard

A. Train separate models for each language from scratch.

B. Use a rule-based system with language-specific sentiment dictionaries.

C. Use an English-only model and translate all inputs to English before analysis.

D. Use a single pretrained multilingual model fine-tuned on combined data from all three languages.

Solution

1. Step 1: Consider resource and accuracy trade-offs

   Training separate models is resource-heavy; rule-based systems lack accuracy; translation adds errors.

2. Step 2: Choose multilingual fine-tuning

   Fine-tuning one multilingual pretrained model on combined data leverages shared knowledge and saves resources.

3. Final Answer:

   Use a single pretrained multilingual model fine-tuned on combined data from all three languages. -> Option D

4. Quick Check:

   Multilingual fine-tuning balances accuracy and efficiency [OK]

Hint: Fine-tune one multilingual model on all languages together [OK]

Common Mistakes:

• Training separate models wastes resources
• Relying on translation reduces accuracy
• Using rule-based methods limits performance