Please enable JavaScript.
Coggle requires JavaScript to display documents.
THE STRUCTURE OF SENTENCES :PENCIL2: D+L2 LECTURE 2 - Coggle Diagram
THE STRUCTURE OF SENTENCES
:PENCIL2:
D+L2 LECTURE 2
What is the point of Grammar?
Implicit knowledge of grammar that we all have.
How do we communicate the scene?
Need to convey three concepts, but there is a
bottleneck
- we can only say one word at a time.
Without something extra, we can’t explain who did what to whom.
Sometime this ordering of information is crucial.
E.g. Mr. Pink shot Mr. White Vs. Mr. White shot Mr. Pink
Two very different outcomes!
Speaker and listener need to reach some agreement about the ordering of concepts.
In English, our common understanding of word order is
Subject-Verb-Object (SVO)
; but in Japanese: its SOV
So Grammar is just a mutually agreed convention about sentence structure to allow multiple concepts to be piped through a linear system (the sentence)
More formally...
The goal of sentence processing is to assign thematic roles to words.
Need to establish who the
AGENT
is and who the
PATIENT / RECIPIENT
is.
In order to assign these thematic roles, the listener determined the syntactic structure of the sentence (
PARSING
)
Parsing
Fundamentally different from word recognition.
Constructive
- not a recognition process.
Allows us to deal with novel sentences in a meaningful way (unlike most novel words)
“Twas brillig, and the slithy toves. Did gyre and gimble in the wabe: All mimsy were the borogoves, And the mome raths outgrabe.” - Jabberwocky, Lewis Carroll, 1872
Nonsense words but can still determine grammaticality.
Can put
meaningless words
into
meaningful sentence structure.
A simple theory of grammar
Phrase structure grammar:
Grammar:
Sentence --> Noun Phrase + Verb Phrase
Noun Phrase --> Proper Noun
Noun Phrase --> Determiner + Noun
Noun Phrase --> Noun Phrase + Prepositional Phrase
Prepositional Phrase --> Preposition + Noun Phrase
Verb Phrase --> Verb
Verb Phrase --> Verb + Adjective
Verb Phrase --> Verb + Sentence Complement (incomplete sentence)
Lexicon:
Determiner = [the, a, every]
Noun = [man, woman, book, hill, telescope]
Proper Noun = [Rhett, Scarlet]
Preposition = [on, with]
Verb = [saw, put, open, shot, kissed]
Adjective = [pretty, ancient, tall]
Syntactic Tree Structure
"The girl knew the language was ancient."
"
The girl
" = Noun Phrase --> Determiner + Noun
-"
Knew
" = Verb Phrase --> Verb
-- Sentence Complement --
↓ ↓ ↓ (makes sense on its own, but not in this context) ↓ ↓ ↓
"
The language
" = Noun Phrase --> Determiner + Noun
"
Was ancient.
" = Verb Phrase --> Verb + Adjective
Ambiguity in the press...
Syntactic ambiguity is rife in language.
Lots of different ways of interpreting the same sentence.
E.g. Clickbait!
Headlines:
“Hospitals named after sandwiches kill five.”
Syntactic Ambiguity
Occurs when a sentence may be generated in more than one way by phrase structure rules.
Groucho Marx: “One morning I shot an elephant in my pyjamas. How he got in my pyjamas I’ll never know.”
Works because one interpretation is implausible.
But how about: “The girl hit the man with the umbrella.”
Did the man have the umbrella or did the girl?
Attachment Ambiguity
Prepositional Phrases may attach to Noun Phrases and Verb Phrases.
How do we decide?
Local Ambiguity
Some ambiguity is global - unresolved at the end of the sentence.
But much ambiguity is temporary (local)
“And I can’t believe that I’m sharing a kebab with the most beautiful girl I have ever seen | with a kebab.”
“The athlete realised his goals | were unattainable.”
Assume that the sentence would end before the |.
Do we begin by ‘constructing’ one representation, then reanalyse where necessary?
Incremental parsing
Or wait until the end of the sentence where ambiguity is reduced/resolved?
How does the human parser work?
What do we do if there is more than one interpretation?
“The girl hit the man with the umbrella.”
Single parse vs. all possible interpretations.
Incrementality
“The athlete realised his goals were unattainable.”
How much is interpreted incrementally (word-by-word)?
In either case, what information is used to select the appropriate parse of a sentence?
Just the knowledge of possible syntactic roles?
Or other factors such as plausibility within sentence context?
Tyler & Marslen-Wilson (1977)
Do we make use of prior context halfway through a sentence when we are trying to determine the syntactic roles of particular words?
Examples:
"If you walk near the runway, landing planes..."
--> IS (inappropriate) ARE (appropriate)
"If you've been trained as a pilot, landing planes..."
--> IS (appropriate) ARE (inappropriate)
Predictions...
Non-incremental model
- No difference
Incremental model
- Faster responses for contextually appropriate continuations.
The Garden Path Theory
Frazier & Fodor (1978); Frazier & Rayner (1982)
Incrementality
Each word is assigned a role in the sentence as it is recognised.
Autonomous
The parser doesn't make use of the plausibility of sentence context.
Serial
Only one possibility considered at any one time.
Innate
Strategies for choosing most likely parse are generic across all languages.
Maze with lots of different options
Go down one path, dead end, turn back and find another route.
Ambiguity resolution strategies
Need to choose which alternative route to take.
Need to choose which tree structure works the best.
Minimal Attachment
Adopt the analysis that requires the fewest nodes.
AKA the fewest branches on the syntactic tree.
Go for the simplest option.
But what if both options are equally complex?
Late Closure
If two tree structures are equally complex, attach material into the most recently constructed phrase marker/part of the tree structure.
-
Attach to the most recent construction.