Difference between revisions of "Syntactic labels"

From Apertium
Jump to navigation Jump to search
Line 1: Line 1:
In some language pairs, shallow syntax tags are used to improve disambiguation, or allow tighter rules to be written. For example, disambiguating verb phrase co-ordinators from noun phrase co-ordinators lets you write rules to merge two co-ordinated NPs.
+
In some language pairs, syntactic function labels are used to improve disambiguation, or allow tighter transfer rules to be written. For example, disambiguating verb phrase co-ordinators from noun phrase co-ordinators lets you write transfer rules to merge two co-ordinated NPs.
  +
  +
Apertium processes left-to-right longest match, so if we have the following sentence,
  +
  +
* John kicked the ball and Mary caught it.
  +
  +
And we have a rule for {{sc|noun cc noun}}, then we will get the following analysis,
  +
  +
* John kicked the [ball] and [Mary]
  +
  +
But if we can tag the conjunction as being a global conjunction, then we can avoid lumping the subject of the second sentence with the object of the first sentence.
  +
  +
* [John kicked the ball] and [Mary caught it]
   
 
==Example==
 
==Example==

Revision as of 14:12, 3 December 2009

In some language pairs, syntactic function labels are used to improve disambiguation, or allow tighter transfer rules to be written. For example, disambiguating verb phrase co-ordinators from noun phrase co-ordinators lets you write transfer rules to merge two co-ordinated NPs.

Apertium processes left-to-right longest match, so if we have the following sentence,

  • John kicked the ball and Mary caught it.

And we have a rule for noun cc noun, then we will get the following analysis,

  • John kicked the [ball] and [Mary]

But if we can tag the conjunction as being a global conjunction, then we can avoid lumping the subject of the second sentence with the object of the first sentence.

  • [John kicked the ball] and [Mary caught it]

Example

$ echo "Gud talaði øll hesi orð og segði Hann:" | lt-proc fo-is.automorf.bin | cg-proc fo-is.rlx.bin 
  ^Gud/Gud<np><al><m><sg><acc><@OBJ→>/Gud<np><al><m><sg><nom><@SUBJ→>$ 
  ^talaði/tala<vblex><past><p2><sg><@+FMAINV>/tala<vblex><past><p3><sg><@+FMAINV>$ 
  ^øll/allur<prn><qnt><nt><pl><acc><@←OBJ>$ 
  ^hesi/hesin<prn><dem><nt><pl><acc><@←OBJ>$ 
  ^orð/orð<n><nt><sg><acc><ind><@←OBJ>/orð<n><nt><pl><acc><ind><@←OBJ>$ 
  ^og/og<cnjcoo><@CNP>/og<cnjsub><@CVP>$ 
  ^segði/siga<vblex><past><p3><sg><@+FMAINV>$ 
  ^Hann/Prnpers<prn><p3><m><sg><nom><@←SUBJ>$^:/:<sent>$

Here, we could for example have a rule that moves subjects of a finite main verb that are to the right, to the left. e.g. @+FMAINV @←SUBJ to @→SUBJ @+FMAINV as is the order in English.

Standard syntax tags

These are the uniform tags used in many Giellatekno projects.

Tag Description
@←SUBJ Subject, head verb to the left
@SUBJ→ Subject, head verb to the right
@←OBJ Direct object, head verb to the left
@OBJ→ Direct object, head verb to the right
@←IOBJ Indirect object, head verb to the left
@IOBJ→ Indirect object, head verb to the right
@→N Noun modifier, head noun to the right
@N← Noun modifier, head noun to the left
@→A Adjective modifier, head noun to the right
@A← Adjective modifier, head noun to the left
@IM
@SPRED Subject predicate
@←SPRED Subject predicate, head verb
@SPRED→
@OPRED
@←OPRED
@OPRED→
@+FAUXV Finite auxiliary verb
@-FAUXV Non-finite auxiliary verb
@+FMAINV Finite main verb
@-FMAINV Non-finite main verb
@-FSUBJ→
@-F←OBJ
@-FOBJ→
@SPRED←OBJ
@-FADVL
@←ADVL Adverbial modifier, head to the left
@ADVL→ Adverbial modifier, head to the right
@ADVL Adverbial modifier
@P← Complement of a preposition
@CNP Local conjunction or subjunction
@CVP Conjunction or subjunction that joins finite-verb phrases
@→CS
@CNP-VP Ambiguous co-ordinator
@APP
@IMV Infinite main verb
@ICL-ADVL Non-finite subclause ...
@ICL-AUX← "right" argument of auxiliary (?)
@ICL-OBJ Non-finite subclause ...
@ICL-STA Non-finite subclause ...
@HNOUN Noun phrase fragment
@X No analysis

External links