ErgTreebankingGuidelines

Heuristics for efficient treebanking

Top-down

• Choose the construction that spans the whole sentence
  • Typically SUBJH
  • Typically not one of the FRAG* rules

Bottom-up

• Disambiguate lexical entries early, to reduce remaining ambiguity

Technical choices

Complex proper names

Titles

• |Mr. Browne|
  • Choose NP-TITLE-CMPND, not APPOS

Capitalized words in name

• treat as parts of name, not ordinary words
• |Rolls-Royce Motor Cars Inc.|
  • |Motor Cars|
    • NP_NAME_CMPND, not NOUN_N_CMPND
  • |Rolls-Royce|
    • Choose multi-word entry when available
  • |Rolls-Royce Motor Cars|
    • NP_NAME_CMPND
  • Attach |Inc.| with NADJ_RR

Profession modifier

• treat as appositive
• |Howard Mosher, president and CEO|
  • First combine |Howard Mosher|
  • Then combine it with |president and CEO| using APPOS_NBAR

Native names preferred when available

• Company names
  • |Rolls-Royce|
    • Choose n_-_pn_le, not NP_NAME_CMPND
• Country names
  • |U.S.|
    • Choose n_-_c-nm-pd_le, not n_-_pn-gen_le

Proper names and punctuation

• Unknown names
  • |Elianti.|
    • Choose PUNCT_PERIOD_ORULE (period is not part of name)
• Name abbreviations containing periods
  • |U.S.|
    • Choose PUNCT_PERIOD_ORULE if word is at end of sentence

PP attachment

• Choose highest attachment point consistent with meaning
  • |remain steady at 1,200 cars|
    • attach to VP, not to |steady|
  • |reserve a room for Browne|
    • attach to VP, not to |room|
• In copula constructions (with forms of verb "be"), attach PP inside
  • |be payable Feb. 15|
    • First combine |payable| with |Feb. 15| with HADJ_I_UNS
• Complement vs. modifier - choose complement when available
  • |based in Los Angeles|
    • Choose HCOMP, not HADJ_I_UNS
• PP modifier inserted between verb and its complement NP
  • |publish in statements the names of insiders|
    • First combine |publish| with |in statements| using VMOD_I

Temporal modifiers

• When precede VP, attach to subject NP
  • |the maker last year sold cars|
    • attach |last year| to |maker|
• Treat as modifiers, pumping temporal NP to a PP
  • |last year|
    • Choose NPADV, not ADJN
  • |Feb. 15|
    • Combine with HSPECHC, then choose NPADV
• Complex phrases
  • |early next year|
    • Combine |early| with |next year| using NADJ_RR

Complex compound nouns

• Choose bracketing with intended sense
  • |luxury auto maker|
    • first combine |luxury| with |auto|
• When intended bracketing is not clear, group from right to left
  • |airline ticket counter|
    • first combine |ticket| with |counter|

Coordination

• Nominal phrases
  • Choose N_COORD_TOP_2, not N_COORD_TOP_3 when given the choice
• Sentence-initial conjunction - treat as incomplete coordination of clauses
  • |But Abrams arrived early.|
    • Combine |But| with |Abrams arrived early.| with HMARK_CL

Passive verb vs. adjective

• Choose verb if the meaning is agentive; otherwise choose adjective
  • |A date hasn't been set|
    • For |set|, choose v_np*_le, not aj_-_i_le

Punctuation

• Paired commas marking off a modifier: choose "paired" rule (-PR suffix)
  • |Bell, based in Los Angeles|
    • Choose NADJ_RC_PR to combine modifier phrase with |Bell|

Adverbs

• Negation - always attach |not| to preceding auxiliary if possible
  • |did not meet|
    • First combine |did| with |not| using HCOMP
• Other adverbs between auxiliary and main VP - attach adverb to following VP
  • |can really sing|
    • First combine |really| with |sing| using ADJH_S
• Sentence-initial - Prefer attachment without extraction when possible
  • |Apparently the commission met|
    • Choose ADJ_S, not FILLHEAD_NON_WH_IG

Measure phrases

• Degree modifiers - combine with the number word
  • |about 25 % of them|
    • First combine |about| with |25| using HSPECHC
    • Combine |%| with |of them| using HCOMP
• Dollar amounts - treat the symbol |$| as the head (the unit of measure)
  • |$ 80 billion|
    • Combine |$| with |80 billion| using MEAS_NP_SYMB

Quotations with explicit attribution

• treat as extraction from 'saying' verb
  • |They arrived, Browne said.|
    • Combine |They arrived,| with |Browne said.| using FILLHEAD_NON_WH

Partitive NPs

• First pump determiner to noun, and treat of-PP as complement
  • |some of the books|
    • Combine |some| with |of the books| using HCOMP
• For |all|, |not all|, |both|, and |half|, treat following NP as complement
  • |not all those who wrote|
    • For |not all|, choose native entry n_np_mc-neg_le
    • Combine |not all| with |those who wrote| using HCOMP

Modification in noun phrases

• Modifiers to the right of the head noun are always attached _before_
  • any modifiers to the left
  • |important changes by the SEC|
    • First combine |changes| with |by the SEC| using NADJ_RR

Notes from Tomar meeting

1. Where lexical ambiguity is hard to decide (e.g. even-deg vs even-conj), choose based on frequency in redwoods/deepbank

2. Disprefer modifier attachment to semantically vacuous heads e.g. attach adverbs to hiring..., not be hiring...

3. For there-copula:

   1. Avoid double-object choice and avoid modification of there-cop
   2. Also prefer low attachment of modifier after obj NP
   3. Accept extraction of PP for there-cop as is

4. When choice of verb-particle or verb-mod as in go away, if you can modify the `particle' as in go far away, it is not verb-particle.

5. When choice of spr-hd or mod-hd for Adv-Adj, choose mod-hd

6. Avoid adv-add except for not

7. When WH-Q of form NP-be-NP [EMB: guessing this is choose subj-head; Dan please confirm]

8. For complement of saying, if there's a main clause option for the quoted material choose it:

   • |"Who did Kim hire" asked Mary| not |*Who Kim hired, asked Mary|

9. No free relatives

10. Attach three-dot punct as low as possible

11. Reject ellipsis

12. For ndash between clauses, use run-on

13. For degree specifiers, when there's a choice, take the shortest lexent type name

14. Attach subord clause high [EMB: subordinate clauses are understood as clauses with all arguments overt; do not include in+order+to purposives, etc.]