If trees are useful, how do we get them?

Parsing is the process of taking a string and a grammar and returning one or more parse trees for that string

Analogous to running a finite-state transducer over a tape

• However, since CFGs are more powerful
  • That is, there are languages we can capture with CFGs that we can’t capture with finite-state methods
• The parsing process is likewise more complicated
  • As we'll see in a few days

By grammar, or syntax, we have in mind the kind of implicit knowledge of your native language that you had mastered by the time you were 3 years old without explicit instruction

Not the kind of stuff you were later taught in “grammar” school

• At least not in English-speaking countries :-)
• Indeed some EFL teaching involves something much closer to what we have in mind here

Refers to the way words can be arranged in a given language

Grammars (and parsing) are key components in many applications

• Grammar checkers
• Dialogue management
• Question answering
• Information extraction
• Machine translation

There's a useful (traditional) contrast between two perspectives on this topic:

paradigmatic

What's interchangeable with what?

• words, phrases, . . .

syntagmatic

What co-occurs with what?

• ordering (before/after)
• marking (prefixes/suffixes)

Key notions that we’ll cover

• Categories (paradigmatic)
• Constituency (syntagmatic)
• Heads (syntagmatic)

Key formalism

• Context-free grammars

Groups of words can be shown to act as single units, called constituents

In a given language, these units form coherent classes that behave in similar ways, with respect to

External behavior

How they relate to other units in the language

• We can say that in English, noun phrases can come before verbs

Internal structure

We can describe an internal structure for the class

• This might involve disjunctions of somewhat unlike sub-classes to do this
• For example, English noun phrases can consist of a pronoun, a proper noun, or a complex phrase including a common noun

We can observe some commonality over the behaviour of the following English phrases:

• they
• Cambodia
• my aunt's pen
• the reason I can't stay
• taking another look at Moby Dick
• three french hens

One piece of evidence is that they can all precede verbs

• That is, occur in a frame such as

  • ______________ surprised him.

• This is external, paradigmatic evidence

Internal, syntagmatic, evidence would be, for example, to observe that combining

• determiners  such as "my" or "three"
• qualifiers  such as "aunt's" or "french"
• common nouns  such as "pen" or "hens"

usually results in a coherent phrase, which can fit in the above frame

The internal structure of NPs varies from language to language:

• English  these three expensive books (Dem Num Adj Noun)
• French  ces trois livres chers (Dem Num Noun Adj)
• Chinese  這 三 本 昂貴 的 書 (Dem Num clf Adj part Noun)
• Japanese  これ ら の 三 つ の 高価 な 本 (Dem part pos Num clf pos Adj part Noun)
• Thai  hnạngsụu rākhā phæng sām lèm (Noun Adj Num clf)

There’s nothing easy or obvious about how we come up with

• the 'right' set of constituents
• the rules that govern how they combine

That’s one of reasons there are so many different theories of grammar and competing analyses of the same data

The approach we'll explore isn't exactly "cutting-edge"

• But it's a good compromise between simplicity and adequacy
• And the technology required to support it is a good introduction to what's needed for most other approaches

Also known as phrase structure grammars

• And Backus-Naur form is a standardised approach to notating CFGs

Making explicit the restrictions on rewriting we started with earlier, a CFG consists of

Terminals

or terminal symbols: words (for now)

Non-terminals

or non-terminal symbols: Names for constituents in a language

• E.g., NP (noun phrase), VP (verb phrase), V (verb), S (sentence)

Rules

or productions, each of which is a pair of

a left-hand side

a single non-terminal

a right-hand side

a sequence of any number of terminals and non-terminals

• For example:

  VP → V NP

Distinguished symbol

One of the non-terminals

• The starting point for all analyses
• Usually S

Some overly-simple rules for noun phrases:

NP → Det Nominal
NP → ProperNoun
Nominal → Noun | Nominal Noun

These rules describe two kinds of NPs:

• One that consists of a determiner followed by a nominal
• And another that says that proper names are NPs.

The third rule illustrates:

• A disjunction
  • Two kinds of nominals
  • Not strictly speaking a rule
  • Rather a shorthand notation for two rules
• A recursive definition
  • Same non-terminal on the right and left-side of the rule
  • We can see how this works if we consider a noun phrase such as

    a delivery truck repair manual

  • 

Huddleston and Pullum's The Cambridge Grammar of the English Language is 1860 pages long

• !

So we won't cover all of English by a very long way

• Just enough to uncover some key shortcomings of CFGs

We'll look briefly at

• Sentences
• Noun phrases
  • Agreement
• Verb phrases
  • Subcategorisation

Declaratives

A plane left

S → NP VP

Imperatives

Leave!

S → VP

Yes-no questions

Did the plane leave?

S → Aux NP VP

WH questions

When did the plane leave?

S → WH-NP Aux NP VP

We can identify three quite distinct types of noun phrases:

Pronouns

she, he, we, ...

NP -> Pro

Proper Nouns

Edinburgh, Star Wars, the Eiffel Tower, ...

NP -> PropN

Complex noun phrases

the next prime minister after Thatcher

NP -> CNP

Consider the following moderately complicated noun phrase:

the first three morning flights from Denver to Tampa leaving before 10

We'll need something along the lines of the following tree

That big NP is really about flights

• That’s its central criticial noun
• Let’s call that the head of the NP

We can dissect this kind of NP into:

• The constituents that can come before the head
• The constituents that can come after it

Complex noun phrases can start with determiners

CNP → Det CNP

Determiners can be

Simple lexical items

the, this, a, her

Det → Art

(Recursive) possessives

• simple  Robin’s car
• complex  Robin’s youngest child’s toy

Det → PropN 's
Det → CNP 's

Other premodifiers include

• Quantifiers, cardinals, ordinals:

  • every flight

    CNP → Quant CNP

  • three flights

    CNP → Card CNP

  • first flight

    CNP → Ord CNP

• Adjectives and Adjective phrases:

  • • large cars
    • extremely sleepy baby

    CNP → AP CNP
    AP → Adv AP
    AP → Adj

There are constraints we haven't captured on the order of pre-modifiers:

• Between adjectives and quantifiers:

  • every eligible candidate
  • *eligible every candidate

• Between one adjective and another:

  • big red bus
  • ?red big bus

• Following a common linguistic convention, I'm using an initial asterisk to indicate a word sequence which is not in a (natural) language or cannot (should not) be accepted by a formal grammar
• Likewise an initial question mark for a borderline in/out word sequence

Eventually (or even right away), we get to the Nominal

• Including the head, with or without compounding

CNP → Nominal
Nominal → Noun
Nominal → Nominal Noun

The postmodifiers which stack up behind the head may include

• Prepositional phrases:

  flight from Seattle

• Non-finite clauses (gerundive, infinitive):

  • flights arriving before noon
  • first flight to depart

• Relative clauses:

  • flights that serve breakfast
  • people whom the pilot knows

Similar general (recursive) rules to handle these

Nominal → Nominal PP
Nominal → Nominal GerundVP
Nominal → Nominal InfVP
Nominal → Nominal RelClause