1 of 43

Creating a mini resource grammar

Inari Listenmaa

University of Gothenburg

2017-08-16, Riga

2 of 43

Recap on earlier lectures

3 of 43

Computational Syntax, spring 2023

GF resource grammar

Inari Listenmaa

4 of 43

Recap on GF syntax

Functions

abstract:

fun Hello : Noun -> Greeting ;

concrete:

lin Hello x = "hello" ++ x ;

helper functions:

oper glue : Str -> Str -> Str = \x,y -> x ++ y ;

5 of 43

Recap on GF syntax

Records

definition:

lincat Det = {s : Str ; n : Number} ;

giving a value:

lin a_Det = {s = "a" ; n = Sg} ;

accessing a record (projection):

a_Det.s ; -- has value “a”

Tables

definition:

lincat Noun = Number => Str ;

giving a value:

car_N = table {Sg => "car" ;� Pl => "cars" } ;

accessing a table (selection):

car_N ! Sg; -- has value “car”

6 of 43

Difference between records and tables

Records have named fields that can be of any type.

When projecting a field from a record, you must know the exact name:

{s = "hello" ; compl = "world"}.s ;

Tables are associations from key to value. Keys can only be of finite types (param).

When selecting a value from a table, you can use variables, wildcards, computation:

table {Sg => "car" ; Pl => "cars" } ! Sg ;

s = \\x => noun ! x ++ “foo” ++ adj ! x ;

7 of 43

Recap on GF syntax

Tuples

definition:

Str*Str ;

syntactic sugar for:

{p1 : Str ; p2 : Str} ;

giving a value:

<"hello","world">

Parameters

definition:

param Number = Sg | Pl ;�param Case = Nom | Acc | Dat ;

usage: left-hand side of tables; inherent properties

lincat Noun = Number => Str ;�lincat Det = {s : Str ; g : Number} ;

8 of 43

Motivation for RGL

abstract Hello = {� flags startcat = Greeting ;� cat

Greeting ; Recipient ;�� fun� Hello : Recipient -> Greeting ;� World : Recipient ;� Mum : Recipient ;� Friends : Recipient ;

}

concrete HelloEng of Hello = {

lincat

Greeting, Recipient = {s : Str} ;�� lin� Hello rec = {s = "hello" ++ rec.s} ;� World = {s = "world"} ;� Mum = {s = "mum"} ;� Friends = {s = "friends"} ;��}

9 of 43

Motivation for RGL

concrete HelloIce of Hello = {�� lincat � Greeting = {s : Str} ;

Recipient = {s : Str ; n : Number ;� g : Gender} ;�

param� Gender = Fem | Masc | Neutr ;� Number = Sg | Pl ;�

lin� Hello rec = {s = case <rec.g,rec.n> of� <Sg,Masc> => "sæll" ++ rec.s ;� <Sg,_> => "sæl" ++ rec.s ;� <Pl,Masc> => "sælir" ++ rec.s ;� <Pl,_> => "sælar" ++ rec.s } ;

� World = {s = "heimur" ; g=Masc ; n=Sg} ;� Mum = {s = "mamma" ; g=Fem ; n=Sg} ;� Friends = {s = "vinir" ; g=Masc ; n=Pl} ;�}�

10 of 43

Hello with resource grammar

concrete HelloIce of Hello = open ParadigmsIce, SyntaxIce in {�� lincat � Greeting = Utt ; � Recipient = N ;� lin

Hello r = mkGreeting hello_Interj r ; � World = mkN "heimur" ; � Mum = mkN "mamma" ;� Friends = mkN "vinir" ; � oper � hello_Interj = mkInterj "sæll" ; � mkGreeting = ...

concrete HelloKaz of Hello = open ParadigmsKaz, SyntaxKaz in {�� lincat � Greeting = Utt ; � Recipient = N ;� lin

Hello r = mkGreeting hello_Interj r ; � World = mkN "әлем" ; � Mum = mkN "ана" ;� Friends = mkN "достар" ; � oper � hello_Interj = mkInterj "сәлем" ; � mkGreeting = ...

11 of 43

Resource Grammar Library

Translation in general needs semantic predicates
...but syntactic grammar is useful as library
Linguist implements resource grammar library, domain experts can use it in applications

(You can always write application grammars without RGL too!)

12 of 43

Now write your own!

Piece of cake right?

13 of 43

Mini resource grammar

14 of 43

Miniresource in words (1/6)

Sentence & Utterance

Build a (present tense) sentence from a polarity and a clause
Connect two existing sentences with a conjunction
Build a stand-alone utterance from a sentence or a noun phrase

Polarity

Pos, Neg

Tense

Only present (more will come later!)

15 of 43

Miniresource in words (2/6)

Clause

Build a clause from a noun phrase (= the subject) and a verb phrase

Verb phrase

Build a verb phrase by elevating a verb
Build a verb phrase from a two-place verb and a noun phrase (= the object)
Build a verb phrase from an adjective phrase ("big" → "is big")

Verb & Two-place verb

walk_V, arrive_V, love_V2, ...

16 of 43

Miniresource in words (3/6)

Noun phrase

Build a noun phrase from a determiner and a common noun
Build a noun phrase from a pronoun or a proper noun
Connect two existing noun phrases with a conjunction

Common noun

Build a common noun by elevating a noun
Add an adjective phrase to an existing common noun

Noun

house_N, tree_N, ...

17 of 43

Miniresource in words (4/6)

Adjective phrase

Build an adjective phrase by elevating an adjective

Adjective

big_A, small_A, green_A, ...

18 of 43

Miniresource in words (5/6)

Adverb

Build an adverb from a preposition and a noun phrase

Preposition

in_Prep, on_Prep, with_Prep

19 of 43

Miniresource in words (6/6)

Conjunction

and_Conj, or_Conj

Determiner

a_Det, the_Det, every_Det, …

Pronoun

i_Pron, youSg_Pron, they_Pron, …

20 of 43

Where to start?

Lexical categories / Morphology

Easy to concentrate on one thing
Can be reused for the full resource

Syntax

Easier to test sentences from the beginning
Simplified morphology: github.com/inariksit/GFSS2017/tree/master/mini

Both

See next slide!

21 of 43

Test your grammar regularly

22 of 43

Morphology & lexical categories

23 of 43

Nouns: what

What distinctions do I need?

In general / for mini resource

English

Variable number: singular / plural
Inherent animacy (who or which)
Inherent gender (correct referent pronoun)

Finnish

Variable number: singular / plural
Variable case: nominative, genitive and 12 more

What distinctions do I need?

In general / for mini resource

German

Variable number: singular / plural
Inherent gender
Variable case: nominative, genitive, accusative, dative

24 of 43

Nouns: what

What distinctions do I need?

English

Noun : Type = {s : Number => Str} ;
param Number = Sg | Pl ;

Finnish

Noun : Type = {s : Number => Case => Str} ;
param Case = Nom | Gen | … ;

What distinctions do I need?

In general / for mini resource

German

Noun : Type = {s : Number => Case => Str ;� g : Gender } ;
param Gender = Masc | Fem | Neutr ;

25 of 43

Nouns: how

English: Noun : Type = { s : Number => Str } ;

car_N = { s = table { Sg => “car” ; Pl => “cars” } ;

bus_N = { s = table { Sg => “bus” ; Pl => “buses” } ;

baby_N = { s = table { Sg => “baby” ; Pl => “babies” } ;

26 of 43

Nouns: paradigms

oper carNoun : Str -> Noun ;

carNoun car = { s = table { Sg => car ; Pl => car + “s” } ;

oper busNoun : Str -> Noun ;

busNoun bus = { s = table { Sg => bus ; Pl => bus + “es” } ;

oper babyNoun : Str -> Noun ;

babyNoun baby = { s = table { Sg => baby ; Pl => init baby + “ies” } ;

27 of 43

Nouns: smart paradigms

oper regNoun : Str -> Noun = \s -> case s of {

_ + ("s" | "ch") => { s = table { Sg => s ; Pl => s + "es" } ;

_ + ("y") => { s = table { Sg => s ; Pl => init s + "ies" } ;

_ => { s = table { Sg => s ; Pl => s + "s" }

} ;

28 of 43

Smart paradigms: Dutch

regNoun : Str -> Noun = \s -> case s of {

_ + ("a" | "o" | "y" | "u" | "oe" | "é") => mkNoun s (s + "'s") Utr ;

_ + ("oir" | "ion" | "je") => mkNoun s (s + "s") Neutr ;

? + ? + ? + _ +

("el" | "em" | "en" | "er" | "erd" | "aar" | "aard" | "ie") -- unstressed

=> mkNoun s (s + "s") Utr ;

_ + ("i"|"u") => mkNoun s (endCons s + "en") Utr ;

b + v@("aa"|"ee"|"oo"|"uu") + c@? => mkNoun s (b + shortVoc v c + "en") Utr ;

b + ("ei"|"eu"|"oe"|"ou"|"ie"|"ij"|"ui") + ? => mkNoun s (endCons s + "en") Utr ;

_ + "ie" => mkNoun s (s + "ën") Utr ;

b + v@("a"|"e"|"i"|"o"|"u") + c@? => mkNoun s (b + v + c + c + "en") Utr ;

_ => mkNoun s (endCons s + "en") Utr

} ;

29 of 43

Applying smart paradigms

> regNoun "meisje"� s . NF Sg Nom => meisje� s . NF Sg Gen => meisjes� s . NF Pl Nom => meisjes� s . NF Pl Gen => meisjes� g . Neutr

> regNoun "muis"� s . NF Sg Nom => muis� s . NF Sg Gen => muis� s . NF Pl Nom => muizen� s . NF Pl Gen => muizens� g . Utr

(later: how about the nice mkN function?)

30 of 43

Verbs (Italian)

Variable: Person, tense, mood, polarity
Inherent: Which auxiliary it takes for participle
Example: MiniresourceIta.gf#L265

32 of 43

Phrasal categories

“Next step” from noun, adjective and verb:

Noun → CN, NP
Adjective → AP
V, V2 → VP

Morphology is there, how do we use it?

Agreement: select appropriate things from inflection tables

Can we get rid of some distinctions? Do we need to add new ones?

e.g. English case in NP

33 of 43

Miniresource functions:

cat AP ;

fun UseA : A -> AP ;

A to AP: if A has variation (e.g. number & gender) that depends on its eventual head, generally AP keeps that
Remember that AP can be used to form both NPs and VPs!

34 of 43

CN and NP

Miniresource functions:

cat CN ;

fun UseN : N -> CN ;

fun AdjCN : AP -> CN -> CN ;

cat NP ;

fun DetCN : Det -> CN -> NP ;

fun ConjNP : Conjunction -> NP -> NP -> NP ;

N to CN

CN still has variable number
AP’s gender/class/... gets fixed

CN to NP: number gets fixed, case still variable

35 of 43

Miniresource functions:

cat VP ;

fun UseV : V -> VP ;

fun ComplV2 : V2 -> NP -> VP ;

fun CompAP : AP -> VP ;

V to VP: keep subject agreement open
V2 to VP: V2’s object agreement becomes fixed; NP’s case becomes fixed

37 of 43

Clauses

Miniresource functions:

cat S, Cl, Pol ;

fun PredVP : NP -> VP -> Cl ;

fun UsePresCl : Pol -> Cl -> S ;

fun ConjS : Conjunction -> S -> S -> S ;

NP & VP to Cl: VP fixes subject agreement (example)
Cl to S: fixes polarity
How about word order?

Maybe not needed in miniresource, but think where to add the feature!

39 of 43

Is this useful for more than exercise?

40 of 43

Effort

Full resource grammar: months

Mini resource grammar:

days to weeks

RGL

Wide-coverage

Wide-coverage grammar:

days if we have RGL

Mini

41 of 43

Effort

Wide-coverage = mostly allowing chunks in yellow grammar
If we only have structures in mini, can we still parse something from most sentences?

RGL

Wide-coverage

Mini

42 of 43

Conclusion

Writing miniresources is…

not scary at all! It’s just 6 slides of stuff
fun, even for languages you don’t know (well)
educational and useful, prepares for full resource

you’ll have to address many of the hard problems already in mini

1 of 43

2 of 43

3 of 43

4 of 43

5 of 43

6 of 43

7 of 43

8 of 43

9 of 43

10 of 43

11 of 43

12 of 43

13 of 43

14 of 43

15 of 43

16 of 43

17 of 43

18 of 43

19 of 43

20 of 43

21 of 43

22 of 43

23 of 43

24 of 43

25 of 43

26 of 43

27 of 43

28 of 43

29 of 43

30 of 43

31 of 43

32 of 43

33 of 43

34 of 43

35 of 43

36 of 43

37 of 43

38 of 43

39 of 43

40 of 43

41 of 43

42 of 43

43 of 43