Constituency Grammars Flashcards
What does syntax mean?
It is the way words are arranged together
What does syntactic consistency mean?
It is the idea that words can be grouped into single units (e.g. Noun Phrase)
How do we form constituents?
We use evidence from the context of the sentence to group words to form them
What is a constituent?
It is a word or group of words that function as a single unit
How can evidence be encoded?
In rules or grammars
What are some different Grammar types?
Context Free Grammar (CFG)
Dependency Structure Grammar
What are some features of a CFG?
Rules are based on phrasal constituents + phase-structure
Word order is very important
Head terms are embedded into trees making it harder to find
What are some features of a Dependency Structure Grammar?
Rules are based on grammatical dependencies between words
Word order is flexible
(Head → Dependent) approximates the semantic relationship between predicates and arguments
State which of the graphs show a CFG and a Dependency Grammar.
The left is a Dependency Grammar
The right is a CFG
Explain what the graph generated using CFG shows.
The root node, S, is saying we have a sentence, and for this parse tree the sentence will have a noun phrase followed by a verb phrase. We can see that the noun phrase consists of a pronoun, which we can see is ‘I’. We can see that the verb phrase consists of a verb and noun phrase. This continues until we reach all the leaf nodes.
What are the leaf nodes typically in a CFG graph?
They are typically lexical terms (words)
What are dependency grammars based on?
They are based on the subject-object relationship.
What does a context free grammar model?
It models constituent structure
What two things does a CFG have?
A lexicon (of words and symbols)
A set of rules (or productions)
What form are CFG rules equivalent to?
Backus-Naur Form (BNF)
How are CFG rules embedded, and what does it allow them to do?
They are hierarchically embedded, meaning that they can trigger other rules
Explain how the productions work in the image.
Given a left symbol, generate a right set of symbols.
One derivation is where you follow the productions all the way through
The productions can actually be recursive
What are some typical CFG production rules?
Leaf nodes are terminal nodes
Non-terminal nodes define lexical categories (POS)
A node is said to dominate its child nodes
The root node is the start symbol (usually ‘S’)
What do we call sentences that can and cannot be derived from a CFG?
Can be derived are called grammatical, cannot be derived are called ungrammatical
Why is a CFG a generative grammar?
The language is defined by the possible sentences it can generate
What is the problem of mapping sentences to parse trees called?
Syntactic parsing
What does a declarative sentence mean?
We have a subject NP followed by an VP
S → NP VP
What does an imperative level sentence mean?
There is a VP with no subject
S → VP
What does a yes-no question form mean?
There is an auxiliary verb followed by a subject NP and a VP
S → Aux NP VP
What does a wh-subject-question mean?
There is a wh-word followed by a subject NP and a VP
S → Wh-NP VP
What does a wh-non-subject-question mean?
It is where the wh-phrase is not the subject
S → Wh-NP Aux NP VP
What is a wh-non-subject-question an example of?
A long distance dependency - the Wh-NP is far away from the semantically relevant main VP
What can sentences consist of?
One or more clauses
What is a clause?
It is a ‘complete thought’
What is a clause made up of?
Two or more of:
- Subject
- Verb
- Object
- (Subject|Object) Complement
- Adverbial
What are clauses critical for?
Applications such as relation extraction
What do noun phrases consist of?
A head noun and various modifiers
What can determiners be?
Simple lexical terms like a, the, this.
Or they can be more complex with possessive markers (‘s)
What is a nominal?
It is typically a head noun and optional noun modifiers which can occur before or after the head noun
Nominal → Noun | NUM Nominal | Nominal PP | (who|what) VP
What is verb phrase typically?
A VP plus a number of other constituents
VP → VP | Verb NP PP | Verb NP | Verb PP
What are sequential compliments?
These are VP followed by an embedded sentence
VP → Verb S
What do traditional grammars do with verbs?
Subcategorise the verbs into a few categories
What are some verb categories?
Transitive Verbs - object
Intransitive Verbs - no object
Ditransitive Verbs - direct and indirect object
Linking Verbs - links clause subject with complement
What is different about what modern grammars do with verbs?
They can have up to 100 subcategories of verbs. Sets of complements are called the subcategorisation frame
What are coordinations?
These are conjunctions such as and, or, but
VP → NP and NP
Nominal → Nominal and Nominal
S → S and S
VP → VP and VP
What is a treebank?
A treebank is a syntactically annotated corpus. They have tagsets based on linguistic annotation choices from the authoring project
How are long distance dependencies (syntactic movement) encoded in tagsets?
They are encoded using -NONE- markers
How many rules types and words does Treebank 3 have?
17,500 rule types and a million words
What is a lexical head?
A lexical head is the word in a phrase which is grammatically the most important - they can be tricky to define as they are context dependent. In practice, handwritten rules are used guided by statistical analysis
What grammar emphasises lexical features over phrase-structure?
Combinatory Categorical Grammar (CCG) - defines categories and has mappings between lexicon words to categories, can go forward and backwards. Is very powerful
