Exploring Machine Translation Concepts

What is the primary purpose of Machine Translation?

• To automatically convert text from one language to another.

• To analyze the sentiment of text in different languages.

• To create original content in multiple languages.

• To translate spoken language in real-time.

Describe the historical evolution of Machine Translation.

• The historical evolution of Machine Translation includes Rule-Based MT, Statistical MT, and Neural MT.

• Machine Translation relies solely on human translators.

• Machine Translation has only one method: Neural MT.

• Machine Translation was invented in the 21st century.

What are the key differences between statistical and neural approaches in MT?

• Neural approaches are based on linear regression, while statistical methods use decision trees.

• Statistical approaches rely solely on rule-based systems, while neural approaches use traditional algorithms.

• Statistical approaches use probabilistic models and large corpora, while neural approaches leverage deep learning for better context understanding and fluency.

• Statistical methods focus on small datasets, whereas neural methods require minimal data for training.

Explain phrase-based translation in statistical models.

• Phrase-based translation translates entire documents without statistical models.

• Phrase-based translation does not involve any alignment between source and target languages.

• Phrase-based translation uses statistical models to translate segments of text by aligning phrases in source and target languages.

• Phrase-based translation relies solely on word-for-word translation techniques.

What is language modeling in the context of MT?

• Language modeling in MT is the process of predicting word sequences to improve translation fluency and coherence.

• Language modeling is the process of translating text without considering context.

• Language modeling focuses solely on vocabulary expansion.

• Language modeling is only about grammar correction in MT.

Define decoding algorithms and their role in MT.

• Decoding algorithms are processes that convert encoded text into a target language in machine translation, ensuring fluent and accurate translations.

• Decoding algorithms are used to encode text into a target language.

• Decoding algorithms are irrelevant in machine translation processes.

• Decoding algorithms only focus on grammatical structure without considering fluency.

What are the main human evaluation metrics for MT?

• Relevance, coherence, translation speed

• Clarity, complexity, user satisfaction

• Adequacy, fluency, post-editing effort

• Consistency, accuracy, readability

How does the BLEU score measure translation quality?

• The BLEU score measures translation quality by counting the total number of words in the candidate translation.

• The BLEU score measures translation quality by comparing n-grams in the candidate translation to reference translations and calculating precision with a brevity penalty.

• The BLEU score assesses translation quality by analyzing the grammatical structure of the candidate translation.

• The BLEU score evaluates translation quality based on the length of the sentences in the candidate translation.

What is the significance of adequacy and fluency in MT evaluation?

• Adequacy ensures cultural relevance, while fluency ensures adherence to strict formatting rules.

• Adequacy ensures meaning preservation, while fluency ensures naturalness in translation.

• Adequacy is about the speed of translation, while fluency is about the use of technical jargon.

• Adequacy focuses on grammatical correctness, while fluency focuses on word count.

Describe the process of preprocessing parallel corpora.

• The process of preprocessing parallel corpora includes data collection, cleaning, tokenization, alignment, normalization, filtering, encoding, and splitting.

• Translation of texts without alignment

• Only encoding and splitting the data

• Data collection and analysis only