Discourse and Pragmatic Analysis

Discourse Analysis

While semantic analysis handles the meaning of a single sentence, discourse analysis studies meaning across multiple connected sentences — a paragraph, a conversation, or a full document.

Key discourse-level problems:

1. Anaphora resolution (coreference resolution) — determining what a pronoun or noun phrase refers to.

text = "Riya bought a laptop. She said it was expensive."
# "She"  -> refers to "Riya"
# "it"   -> refers to "a laptop"

1. Discourse structure / coherence — how sentences logically connect (cause-effect, contrast, elaboration).

"The exam was postponed. Students were relieved."
        ^                        ^
      cause                   effect (implicit "because of this")

1. Topic tracking — identifying what a multi-sentence passage is "about" as it evolves.

Pragmatic Analysis

Pragmatics is the highest level — it studies meaning in context of real-world use: the speaker's intent, the situation, shared knowledge, tone, and social conventions. The same sentence can carry different pragmatic meaning depending on context.

sentence = "Can you pass the salt?"
# Literal (semantic) meaning : a yes/no question about ability
# Pragmatic meaning          : a polite request to pass the salt

Summary — All Levels Together

A production NLP pipeline (e.g. a chatbot) typically touches every one of these levels, from tokenizing input text all the way to inferring user intent.