How deeply does LLMs in Production cover pragmatics?

Defines context through interpretation and works through prompting and database access, while denying that these create an internal pragmatic system.

Pragmatics

As explained in LLMs in Production

Pragmatic context is contextual information about real-world tasks, expectations, and circumstances that affects how language is interpreted or generated.

A shopping assistant can generate a fluent reply and still recommend the wrong item if it does not know what products exist, where they are sold, or what they cost. Pragmatics is the real-world, task-specific context that helps make a response appropriate—not merely well formed. section 2.4 section 5.3.1

Pragmatics is one part of language alongside phonetics, syntax, semantics, and morphology. It matters when words alone do not settle what someone means: “I need food” does not necessarily signal immediate physical danger. Embeddings can approximate semantic relationships from word usage, but those relationships do not supply the circumstances behind an utterance. section 2.1.1 section 2.2.5

For an LLM, supply relevant context rather than assuming fluent output means the model knows the situation:

Put task instructions and relevant facts in the prompt, or retrieve product, store, and price information from a database at inference time. This can improve the response without creating a pragmatic system inside the model. section 2.4
For changing information such as current events, keep retrieved information up to date or give the model access to tools; a well-written prompt cannot provide facts the model cannot access. section 7.4
During training, prompt tuning can provide task context while reusing a largely frozen model. It offers an alternative to retraining or full fine-tuning when adapting a model to a new task. section 5.3.1

How LLMs in Production covers it

liveBook manning.com

Read LLMs in Production on liveBook Get the book at manning.com

Related concepts

natural languagepart of · Natural language is a human language that develops through social use and is used for communication, in spoken or written form.
prompt engineeringused by · Prompt engineering is the deliberate design and refinement of instructions, context, examples, or other inputs to guide an AI model’s behavior or outputs.
retrieval-augmented generationused by · Retrieval-augmented generation is a technique that retrieves relevant information from external sources and supplies it as context to a generative model when producing an output.
embeddingcontrasts with · A numerical vector representation of an item or input that encodes features or relationships useful for computational tasks.

This explainer was generated by Manning's knowledge graph from the text of LLMs in Production. The section links point to the parts of the book it draws on.