Updated Stanford to UD converter

  * Note for optimal behavior, morph column should include 'person' if a token is the head of a person entity
  * Currently assumes lemmas are available for copula handling (this can probably be fixed)

examples/stan2uni.ini

@@ -1,56 +1,153 @@
 ;;;Configuration file to convert Stanford Typed Dependencies to Universal Dependencies
-;;;Note not all aspects of universal dependencies are covered, especially handling of the 'name' label
+;;;Note not all aspects of universal dependencies are covered, especially handling of the 'flat' label requires NER information assumed below
+
+;handle free relatives such as "an expectation of what to do" (change pcomp to pobj+rcmod)
+func=/prep/;pos=/^W.*/;func=/pcomp/	#1>#3;#3>#2	#2:func=pobj;#1>#2;#2>#3;#3:func=rcmod
 
 ;simple rename
 func=/^acomp/	none	#1:func=xcomp
 func=/^npadvmod/	none	#1:func=nmod:npmod
 func=/^num/	none	#1:func=nummod
 func=/^rcmod/	none	#1:func=acl:relcl
-func=/^vmod/	none	#1:func=nfincl
+func=/^dobj/	none	#1:func=obj
+pos=/DT/&func=/^neg/	none	#1:func=det
+func=/^neg/	none	#1:func=advmod
+func=/preconj/	none	#1:func=cc:preconj
 func=/^quantmod/	none	#1:func=advmod
-func=/^tmod/	none	#1:func=npmod:tmod
+func=/^tmod/	none	#1:func=nmod:tmod
 func=/^predet/	none	#1:func=det:predet
 func=/^possessive/	none	#1:func=case
 func=/^poss$/	none	#1:func=nmod:poss
 func=/^prt/	none	#1:func=compound:prt
+func=/^auxpass/	none	#1:func=aux:pass
+func=/^csubjpass/	none	#1:func=csubj:pass
+func=/^nsubjpass/	none	#1:func=nsubj:pass
 
 ;disambiguate nn
 func=/nn/&pos=/JJ/	none	#1:func=amod
 func=/nn/&pos=/DT/	none	#1:func=det
+
+; Logan's try of flat (example: actor Bob A B C D Smith)
+; First deal with person entities based on 'person' in the morph field
+; step 1	let Bob dominate all ABCD
+; non-sentence-initial
+pos!=/N?NP.*/;pos=/N?NP.*/&func=/nn/;pos=/N?NP.*/&func=/nn/;pos=/N?NP.*/&morph=/(.*person.*)/	#1.#2.*#3.*#4;#4>#2;#4>#3	#2>#3;#3:func=flat
+; sentence-initial
+pos=/N?NP.*/&func=/nn/&position=/first/;pos=/N?NP.*/&func=/nn/;pos=/N?NP.*/&morph=/(.*person.*)/	#1.*#2.*#3;#3>#1;#3>#2	#1>#2;#2:func=flat
+
+; step 2	let anything previously dominated by Smith now by Bob
+pos=/.*/;pos=/N?NP.*/&func=/nn/;pos=/N?NP.*/&morph=/(.*person.*)/	#1.*#2.*#3;#3>#1;#3>#2	#2>#1
+
+; step 3 let Bob dominate Smith and receive func, morph information from Smith
+; head has parent
+func=/.*/;pos=/N?NP.*/&func=/nn/;pos=/N?NP.*/&func=/(.*)/&morph=/(.*person.*)/	#1.*#2.*#3;#3>#2;#1>#3	#1>#2;#2>#3;#2:func=$1;#2:morph=$2;#3:func=flat;#3:morph=_
+; head has no parent
+pos=/N?NP.*/&func=/nn/;pos=/N?NP.*/&func=/(root)/&morph=/(.*person.*)/	#1.*#2;#2>#1	#1>#2;#1:func=$1;#1:morph=$1;#1:head=0;#2:func=flat;#2:morph=_
+
+; step 4 move head of flat fountain to actor if the entity is a person, morph and func info also moved to actor
+func=/nn/;pos=/N?NP.*/&morph=/(.*person.*)/&func=/(.*)/&head=/(.*)/;func=/flat/	#1.#2.*#3;#2>#1;#2>#3	#1>#2;#1>#3;#1:morph=$1;#1:func=$2;#1:head=$3;#2:func=flat;#2:morph=_
+
+; step 4.5: add old flat analysis for non-person
+func=/nn/;func=/nn/;text=/.*/	#1.#2;#3>#2;#3>#1	#1>#2;#2:func=flat
+pos=/N?NP.*|FW/&func=/nn/;pos=/N?NP.*|FW/&func=/(.*)/;text=/.*/	#2>#1;#3>#2	#1>#2;#2:func=flat;#3>#1;#1:func=$1
+pos=/N?NP.*|FW/&func=/nn/;pos=/N?NP.*|FW/&func=/(root)/	#2>#1	#1>#2;#2:func=flat;#1:head=0;#1:func=$1
+
+; step 5 cleaning: whatever dominated by flat will be by its parent
+; precedence shouldn't matter for this step, since flat never dominates anything
+;func=/.*/;func=/.*/;func=/flat/	#1.*#2.*#3;#3>#1;#2>#3	#2>#1
+func=/.*/;func=/.*/;func=/flat/	#3>#1;#2>#3	#2>#1
+
+; all rest nn become compound
 func=/nn/	none	#1:func=compound
-;handle names HERE
-;The following undergenerates, probably best to leave off: 
-;text=/^(Mr|Mrs|Ms|Dr|Prof(essor)?|Doctor|King|Queen|Prince(ss)?|Attorney|Governer)\.?/&func=/^nn/;pos=/^NN?P/	#1.#2;#2>#1	#1>#2;#2:func=name
+
 
 ;to-infinitive
 text=/^[Tt]o$/&func=/aux/	none	#1:func=mark
 
 ;adverbial clause modifying a nominal
 pos=/^(N.*|P.*)$/;pos=/[NV].*/&func=/advcl/	#1>#2	#2:func=acl
 
+;clauses (handling vmod)
+pos=/^[NJ].*$/;pos=/^V[VBH]$|V.N|V.G/&func=/vmod/	#1>#2	#2:func=acl
+pos=/^V.*/;pos=/^V.[GN]$/&func=/vmod/	#1>#2	#2:func=advcl
+pos=/^V.*/;pos=/^V.$/&func=/vmod/	#1>#2	#2:func=advcl
+pos=/^V.*/;func=/vmod/;func=/cop/	#1>#2>#3	#2:func=advcl
+
+;reverse fixed
+func=/^mwe$/;func=/(.*)/;func=/.*/	#3>#2>#1;#1.1,3#2	#3>#1;#1>#2;#2:func=fixed;#1:func=$1
+func=/^mwe$/	none	#1:func=fixed
+
+;handle copula
+;protect expletive existential first
+func=/expl/;lemma=/be/	#2>#1	#2:morph=EXIST
+;now convert copula cases
+func=/nsubj|csubj/;lemma=/be/&morph!=/EXIST/&func=/(ccomp|parataxis|advcl|rcmod)/;func=/prep|pcomp/;func=/pobj/;text=/.*/	#2>#1;#2>#3>#4;#5>#2	#4>#1;#4>#2;#4>#3;#3:func=case;#4:func=$1;#2:func=cop;#5>#4
+func=/nsubj|csubj/;lemma=/be/&morph!=/EXIST/&func=/root/;func=/prep|pcomp/;func=/pobj/	#2>#1;#2>#3>#4	#4>#1;#4>#2;#4>#3;#3:func=case;#2:func=cop;#4:func=root;#4:head=0
+;restore existential case
+morph=/EXIST/	none	#1:morph=_
+
+;TODO: handle copula without lemma (for corpora without gold lemmas)
+
 ;prepositional phrases
+;take care of adnominal pcomp, should become acl (e.g. "mirrors for correcting your pose")
+pos=/^N.*|FW|DT/;func=/prep/;func=/pcomp/	#1>#2>#3	#1>#3;#3>#2;#2:func=mark;#3:func=acl
 ;simple
 func=/.*/;func=/^prep$/;func=/pobj/	#1>#2>#3	#3>#2;#2:func=case;#3:func=nmod;#1>#3
 ;pcomp without mark
 func=/.*/;func=/^prep$/;func=/pcomp/	#1>#2>#3>#4	#3>#2;#2:func=_temp;#3:func=ncmod;#1>#3
 ;pcomp with mark
 func=/.*/;func=/^prep$/;func=/pcomp/;func=/mark/	#1>#2>#3>#4	#3>#2;#2:func=case;#3:func=ncmod;#1>#3
 func=/_temp/	none	#1:func=mark
+;coordinated pobj
+func=/.*/;func=/conj/;func=/pobj/	#1>#2>#3	#2:func=case;#3>#2;#1>#3;#3:func=conj
+
+
+;reverse conjunction 
+func=/.*/;func=/^cc$/;func=/^conj$/	#1>#2;#1>#3	#3>#2
+
+;fix double coordination results in cc attaching to crossed third conjunct
+func=/.*/;func=/^cc$/;func=/^conj$/;func=/^conj$/	#1.*#2.*#3.*#4;#1>#3;#1>#4;#4>#2	#3>#2
+
+
+;prep for verbal predicate with pcomp
+pos=/^V.*|^J.*/;func=/prep/;func=/pcomp/	#1>#2>#3	#1>#3;#3>#2;#2:func=mark;#3:func=advcl
+
+
+;handle obj & xcomp (instead of nsubj of subclause in SD)
+func=/.*/;func=/nsubj/;func=/xcomp/	#1>#3;#3>#2	#1>#2;#2:func=obj
+
+
+;Promote remaining preps, which probably don't have a pobj, to nmod (which subsequently goes to obl if necessary)
+func=/.*/;func=/^prep$/	#1>#2	#2:func=nmod
+
+
+;handle nmod of predicates (should be obl)
+pos=/^(V.*|J.*)$/;func=/nmod/	#1>#2	#2:func=obl
+
+;complex numbers analyzed as compound
+func=/^number/	none	#1:func=compound
+
+;universal POS tags
+cpos=/JJ[RS]?/	none	#1:pos=ADJ
+cpos=/WRB/&func=/advmod|mark|obl/	none	#1:pos=SCONJ
+cpos=/RB[RS]?/&lemma!=/n.?t/	none	#1:pos=ADV
+cpos=/RB/&lemma=/n.?t/	none	#1:pos=PART
+cpos=/UH/	none	#1:pos=INTJ
+cpos=/CC/	none	#1:pos=CCONJ
+cpos=/CD/	none	#1:pos=NUM
+cpos=/NNS?/	none	#1:pos=NOUN
+text=/%/	none	#1:pos=SYM
+cpos=/NNPS?/	none	#1:pos=PROPN
+cpos=/V.*/	none	#1:pos=VERB
+func=/.*aux.*|cop/	none	#1:pos=AUX
+cpos=/IN|RP|TO/&func!=/mark/	none	#1:pos=ADP
+func=/mark/	none	#1:pos=SCONJ
+func=/(pre)?det/	none	#1:pos=DET
+cpos=/POS|TO/&func!=/case/	none	#1:pos=PART
+cpos=/PRP\$?|WP\$?|EX|W?DT/&func!=/det/	none	#1:pos=PRON
+func=/punct/	none	#1:pos=PUNCT
+cpos=/FW|LS/	none	#1:pos=X
+
+;clear morph
+morph=/.*/	none	#1:morph=_
 
-;2word mwes
-;"apart_from", "as_from", "aside_from", "away_from", "close_by", "close_to", "contrary_to", "far_from", "next_to", "near_to", "out_of", "outside_of", "pursuant_to", "regardless_of", "together_with"
-text=/^([Aa]part|[Aa]s|[Aa]side|[Aa]way|[Ff]ar)$/;text=/^from$/	#1.#2	#1>#2;#2:func=mwe
-text=/^[Cc]lose$/;text=/^(by|to)$/	#1.#2	#1>#2;#2:func=mwe
-text=/^[Cc]ontrary$/;text=/^(to)$/	#1.#2	#1>#2;#2:func=mwe
-text=/^([Nn]ext|[Nn]ear|[Pp]ersuant)$/;text=/^(to)$/	#1.#2	#1>#2;#2:func=mwe
-text=/^([Oo]ut|[Oo]utside|[Rr]egardless)$/;text=/^(of)$/	#1.#2	#1>#2;#2:func=mwe
-text=/^[Tt]ogether$/;text=/^with$/	#1.#2	#1>#2;#2:func=mwe
-
-;3word mwes
-;"by_means_of", "in_accordance_with", "in_addition_to", "in_case_of", "in_front_of", "in_lieu_of", "in_place_of", "in_spite_of", "on_account_of", "on_behalf_of", "on_top_of", "with_regard_to", "with_respect_to"
-text=/^[Bb]y$/;text=/^means$/;text=/^of$/	#1.#2.#3	#1>#2>#3;#2:func=mwe;#3:func=mwe
-text=/^[Ii]n$/;text=/^accordance$/;text=/^with$/	#1.#2.#3	#1>#2>#3;#2:func=mwe;#3:func=mwe
-text=/^[Ii]n$/;text=/^addition$/;text=/^to$/	#1.#2.#3	#1>#2>#3;#2:func=mwe;#3:func=mwe
-text=/^[Ii]n$/;text=/^(case|front|lieu|place|spite)$/;text=/^of$/	#1.#2.#3	#1>#2>#3;#2:func=mwe;#3:func=mwe
-text=/^[Oo]n$/;text=/^(account|behalf|top)$/;text=/^of$/	#1.#2.#3	#1>#2>#3;#2:func=mwe;#3:func=mwe
-text=/^[Ww]ith$/;text=/^(regard|respect)$/;text=/^to$/	#1.#2.#3	#1>#2>#3;#2:func=mwe;#3:func=mwe