What primary class of language tasks is the Transformer architecture widely used to power in modern NLP systems?
Select an answer to reveal the explanation.
Short Explanation and Infographic
Let's keep this straight. Transformers blew up because they're great at sequences—translation, summarization, chatty text generation, the works. Attention lets every token look at every other token without the old RNN bottleneck. Your boss says "we need multilingual translation and a drafting assistant"—that's Transformer country. Exam trap: don't pick generic regression, photo classification, or clustering just because those are ML words you know. Those aren't the classic NLP story. Remember: Transformer = modern sequence NLP powerhouse, especially generation and translation. Nice and clean.
Full explanation below image
Full Explanation
The Transformer architecture, introduced with multi-head self-attention and position-aware feed-forward blocks, became the dominant backbone for large-scale natural language processing. Encoder–decoder Transformers popularized high-quality neural machine translation; decoder-only and encoder-only variants scale to language modeling, summarization, question answering, dialogue, and many other sequence-to-sequence or sequence-labeling tasks. In exam-oriented terms, the architecture is commonly associated with text generation and automatic translation—and more broadly with modeling token sequences where long-range dependencies matter.
Self-attention computes interactions among all positions in a sequence (within practical length limits), which improves parallelization during training compared with recurrent networks and captures contextual relationships efficiently. That inductive bias matches language, where meaning depends on distant words and flexible word order patterns. Production systems for translation services, code assistants, and generative applications rest on these ideas.
Selecting pure tabular regression as the main purpose is incorrect: while Transformers can be adapted to many modalities, their celebrated NLP role is not classical business-metric regression on flat features. Whole-image classification is a computer-vision task; although Vision Transformers apply attention to image patches, that is a later cross-domain extension, not the original NLP framing of the question. Unsupervised clustering of numeric rows likewise misses the sequence-modeling and generation focus.
For study purposes, connect Transformer → attention over sequences → translation and generation (plus related NLP tasks). Know that "Transformer" names an architecture family, not a single product, and that training objectives (masked language modeling, causal language modeling, translation pairs) shape what a given model does well. Best practice in systems work includes managing context length, tokenization, and inference cost—but the conceptual answer remains: Transformers excel at modern sequence NLP workloads such as translation and text generation.