Discourse analysis + AI

Question 1

Discourse

Answer 1

sequence of connected utterances

Question 2

Narrative

Answer 2

Narrative = type of discourse
Two or more events that have a temporal connection (time or sequential) and logical connection

Question 3

OCERC

Narrative structure

Answer 3

1. Orientation - who, what, when where, sets scene for the hearer
2. Complication - then what happened? Series of events that make up the story
3. Evaluation - so what? Speakers attitude to the story, giving it significance
4. Resolution - what finally happened? Result of the narrated event/s
5. Coda - what does that mean to us? Returns the perspective to the present moment

Question 4

REDL

Narrative cohesion

Answer 4

Requires tactics to establish relationships between parts of the narrative
–> reduce cognitive burden

• reference
• ellipsis
• discourse conjunction
• lexical cohesion

Question 5

Reference

Answer 5

participants/circumstance is introduced at one place in narrative then serves as a referent point
- refers backs to it
- example - pronoun ‘it’ to refer to noun, 3rd person pronouns for people names (he,she, him, them)
- demonstratives - these, those, this, that

Question 6

Ellipsis

Answer 6

the omission of certain constituents that are grammatically required
- recoverable from context
- eg. pronoun as a noun phrase - ‘they hit the fence. didn’t break it though.’
- eg. verb phrase recoverable - ‘get ice cream’ ‘we couldve []’

Question 7

Discourse conjunction

Answer 7

formal markers relates a sentence to a previous sentence
- temporarily in time
- or logically with cause and effect relationship
- and, anyway, then, now, but

Question 8

lexical cohesion

Answer 8

repeating when not necessary
- same word or semantically similar word
- ‘walked to the phone and the phone rang.’

Question 9

Labov

Reportable event

Answer 9

= one which justifies the automatic reassignment of speaker role to the narrator
- justify the speaker continuing to speak

Question 10

AI, large language models

Answer 10

The distributional hypothesis
- Computers can learn about (say) colours by looking at
where they occur in text (their distribution)
LLMs
- arranged like a series of layers and each layer adds information to the words in the input