Improved Parser for Simple Croatian Sentences

1
Improved Parser forSimple Croatian Sentences

• Kristina Vuckovic, Božo Bekavac, Zdravko Dovedan
• University of Zagreb, Faculty of Humanities and
  Social Sciences
• kvuckovi, bbekavac,zdovedan_at_ffzg.hr
• NooJ2010 Komotini, Greece2010-05-28

2
Problems
Last Years
Solved

• if there is a nominal predicate
• if the ltNPgt node is a coordination made of 2 or
  more ltNPgts of different gender

• ex.
• The child was very smart.
• -gt The child was
• -gtwas very smart.

• ex.
• The boy and a girl were very smart.
• -gtThe boy and a girl gtltNPpmgt

3
Nominal Predicate

• simple ltNPgt
• single noun
• He is a boy.
• complex ltNPgt
• single noun any number of adjectives, pronouns,
  numerals proceeding and agreeing with the noun in
  number, gender and case
• He is my young friend.
• single pronoun
• He is mine.
• single adjective
• He is young.
• single numeral
• He is the first.
• adverb
• He is great.

redefined ltNPgt lt(PROAM)Ngt lt(PROAM)PROgt lt(PRO
AM)Agt lt(PROA)Mgt
4
(No Transcript)
5
Examples

• Pravda i opraštanje su dva temeljna stupa mira.
• Justice and forgiveness are two main pillars of
  peace.

• Cijene su izuetno povoljne.
• Prices are extremely affordable.
• To je izvrsno.
• This is great.

6
Measures and Explanations
Precision
Recall
72,72
69,56
ltto be,VPgt ltNPNomgtAdjective same form as
ltVPPassivegt Knjiga je napisana The book is
written Napisana knjiga je The written
book is
7
Coordination of multiple ltNPgt nodes

• all ltNPgt are in the same gender
• ltNPfltNPfjabukagt, ltNPfkruškagt i ltNPfšljivagtgt
• an apple, a pear and a plum
• combination of masculine and feminine ltNPgt
• ltNPmltNPfjabukagt, ltNPfkruškagt i ltNPmananasgtgt
• an apple, a pear and a pineapple
• combination of masculine and neutral ltNPgt
• ltNPmltNPmananasgt i ltNPndijetegtgt
• a pineapple and a child
• combination of neutral and feminine ltNPgt
• ltNPmltNPfjabukagt i ltNPndijetegtgt
• an apple and a child

8
Coordination of two or more proper nouns
ltNPm
gt ltNPm
gt ltNPm
gt ltNPm
gt Last
Names only ltNPm
gt su otišli ltNPf
gt su
otišle ltNPm
gt su otišli

• ltNPmZagrebgt i ltNPmDubrovnikgt
• ltNPmZagrebgt i ltNPfBarcelonagt
• ltNPmTin Ujdurgt i ltNPmFilip Kocijangt
• ltNPfEma Ujdurgt i ltNPmFilip Kocijangt
• ltNPmUjdurgt i ltNPmKocijangt
• ltNPfUjdurgt i ltNPfKocijangt
• ltNPfUjdurgt i ltNPmKocijangt

9
(No Transcript)
10
Example of concordance
11
Measures
Precision
Recall
100
90
12
Thank you for your attention.
www.accurat-project.eu

• The research within the project ACCURAT leading
  to these results has received funding from
  theEuropean Union Seventh Framework
  Programme(FP7/2007-2013), grant agreement no
  248347.