Revisions of the GDA Tag Set

• Ver. 0.67 (2002-11-20)
  • ref replaced with cp, which is an IDREFS attribute.
  • pel renamed to sub. The sb attribute can appear either in a grandchild element of an element whose syn values is one of c, a, r, and e, or in a child element with a cp attribute.
  • nc, strict, sloppy attributes introduced. sb="nil" no longer used.
  • The i value of the syn attribute abandoned.
• Ver. 0.66 (2002-08-17)
  • phd attribute abandoned.
  • Tags <period> and <periodp> are added, which represent durations.
• Ver. 0.65 (2001-08-20)
  • Relation identifier ini is substituted by int, because `ini' is reversible and hence the reversed name cannot have the reversed meaning.
  • Relation identifier pat is abandoned and subsumed in obj.
  • Attribute fil deleted and ed added.
  • Section 2 (Basiscs) and Section 3 (Composing Semantic Structure) have been merged, and Section 6 (Syntactic Annotation) and Section 9 (Scopes) have been embedded therein.
  • If an element X has an sce attribute pointing at X or one of its descendants, then the scope contains no elments outside X. So an element pointing at itself by sce (sce="self") has no scope.
  • The term `head node' is replaced with `body node.'
• Ver. 0.62 (2001-02-26)
  • With the syn attribute, delete value u, and replace values f and b with d (dependency). Then rename fc and bc to f and b respectively and accordingly revise their specifications to be a little more permissive.
  • Make d the default value for syn in the intrasentential tags in every language.
  • Make c the default value for syn in the non-intrasentential tags.
• Ver. 0.59 (2000-09-06)
  • Relation identifier ben is more restrited, so that not to but for the sake of in give it to him for the sake of her has sem="ben".
• Ver. 0.58 (2000-07-08)
  • Relation identifier app deleted.
  • Relation identifiers rpt (reciprocal partner), rec (reciprocity), coc (cooccurrence), ccm (circumstance) added.
• Ver. 0.57 (2000-06-02)
  • Let the default value of syn attribute in intrasentential tags be the same in each language. Hence its value for English nominal tags is not f but b.
• Ver. 0.56 (2000-03-26)
  • A sort of entity-relation graph introduced as the intermediate semantic structure associated with GDA-annotated linguistic data.
  • rel attribute replaced by opr (operator) attribute. The value of a sem, opr, or grel attribute is one or more concept strings. A concept string is a sequence of concept identifiers separated by dots, and represents a directed path in the semantic structure, which is a path of nodes and arrows which are instances of the concept identifiers. The initial node of the directed path represented by the sem attribute is equal to the final node of the directed path represented by the opr attribute. This node is the referent node of the element. So the semantics of prepositions must be annotated by opr attribute.
  • Sections on word sense and coreference deleted.
• Ver. 0.55 (2000-01-15)
  • Conceptual terms are called concept identifiers.
  • General attributes resp (responsibility), dtp (description type), dep (dependency), cocu (cuncurrent utterance), and coiu (coinitial utterance) added.
  • General attribute dtrs (daughters) replaced with sbu (sub-utterance).
  • Values e (error) and i (idiosyncracy) added to syn attribute.
  • <span>, <bspan> (begin of span), <espan> (end of span), <docAuthor>, <byline>, <bfo> (backward-affixing forward operator), and <fbo> (forward-affixing backward operator) tags added.
  • <cite> and <quobte> tags deleted.
  • Section 5 is now `Relation Identifiers'
  • who and whm regarded not as special attributes for <q> elements, but as relation identifiers.
  • Relation identifier arg (argument), mod (modifier), met (metonymy), bsc (basis of comparison), sim (similarity), who (addressor), whm (addresee), added.
  • Relation identifier ctp and gra deleted.
  • Composition of relation identifiers R and S will be R.S, because : (colon) is used for XML namespace.
  • Section `Communicative Functions' moved in front of `Reference Types'.
  • Relatin identifiers uba (unbounded argument), plg (plug), txx (taxonomic extension), in (taxonomic membership), stx (situational extension), sit (situation), bsc (basis of comparison), ilc (internal location), rpt (reciprocal partner), rcp (reciprocity), ctl (control), tnc (reverse of cnt) added.

    <np syn="c" rel="agt.rcp">Tom and Mary</np> got married.

    mix <np syn="c" rel="pat.rcp">sugar and salt</np>

    mix sugar <ad sem="pat.rpt">with</ad> salt

    talk <ad sem="agt.rpt">with</ad> Tom

    <su syn="bc">Let <np rel="ctl.agt">him</np> go</su>

  • Attribute acp (ancestry copy) added. It entails a copy of the pointed ancester element minus the element with the attribute. The pel attribute may be used to encode substitutions in the copy.

    <su id="S" syn="bc">
      <persnamep id="T">Tom</persnamep>
      lives in
      <np syn="fc">
        a
        <ajp id="B">bigger</ajp>
        house
        <adp dep="B" acp="S" rel="cmp">
          than
          <persnamep pel="T">Mary</persnamep>
        </adp>
      </np>.
    </su>
    

  • Operator `-' to make compound relation identifiers entails a copy of the first argument of the relation identifier preceding `-'.

    Tom is taller <ad sem="cmp-obj">than</ad> Mary.

    a dog bigger <ad sem="cmp-obj">than</ad> this

    Tom knows Mary better <ad sem="cmp-sbj">than</ad> Bob.

  • Attributes param and parel deleted and grel added. The grel attribute associates the element with the nearist H-commanding nominal head element.
  • A relative clause is regarded as a <vp> element headed by the main verb, even when there is a WH complementizer. So a relative clause with a WH complementizer is treated similarly to a topicalized sentence. The association between the head noun and the relative pronoun may be encoded by grel attribute or relational attribute. That between the gap in the relative clause and the WH complementizer should be encoded by relational attribute if necessary.
    • The head noun and the gap may be directly associated where the WH complementizer is absent.

      <np>
        <n>people</n>
        <vp syn="bc">
          <np>I</np>
          want to
          <v grel="obj">meet</v>
        </vp>
      </np>

    • rel="uba" means that `whom' does not semantically depends on `want'.

      <np>
        <n>people</n>
        <vp syn="bc">
          <np id="X" rel="uba" grel="eq">whom</n>
          <np>I</np>
          want to
          <v obj="X">meet</v>
        </vp>
      </np>

    • grel cannot be used if `work' is an element.

      <np syn="bc">
        the
        <n>painter</n>
        <vp>
          <np><ad grel="eq">whose</ad> work</np>
          surprised
          <np>me</np>
        </vp>
      </np>

    • plg cancels uba, so that `work' is associated with `whom' via ben.

      <np>
        the
        <n>man</n>
        <vp>
          <adp id="X" rel="uba.ben">for <n grel="eq">whom</n></adp>
          <np>I</np>
          can
          <v plg="X">work</v>
        </vp>
      </np>

    • Parasitic gaps are not distinguished from normal gaps.

      <np>
        the
        <n>book</n>
        <vp syn="bc">
          <np id="X" rel="uba" grel="eq">which</np>
          <np>I</np>
          have
          <v obj="X">lost</v>
          <adp>before <v obj="X">reading</v></adp>
        </vp>
      </np>

  • Treatment of complicated parentheticals added.
  • <linkdv/> through <linkio/> tags deleted.
  • Relation identifier rec (recipient) replaced with gol (goal).
• Ver. 0.48 (1999-06-25)
  • relational term replaced with relation identifier.
  • The combination operator _ replaced with blank.
  • Relation identifiers stx (stative or situation extension) and sta (state or situation) added.
  • sps attribute removed, to use sce instead. No default reading is assumed for absense of sce, so that each specific scoping must be explicitly marked up.
• Ver. 0.32 (1999-02-11)
  • rel="uba" (unbounded argument) added.
  • Communicative functions revised, adding fcom attribute.
• Ver. 0.31 (1999-01-25)
  • syn="f" (syn="b") now allows each child element to depend on ANY head constituent to its right (left).
• Ver. 0.30 (1999-01-17)
  • rel="bas" (basis), rel="cev", and rel="mat" (material) added.
• Ver. 0.29 (1999-01-04)
  • Referential indices top, self, fwd, and bwd added.
  • syn="fp" and syn="bp" deleted, because <fo> and <bo> are enough.
  • `3. Tags' and `4. Syntactic Tagging' exchanged.
• Ver. 0.28 (1998-12-22)
  • syn="fc" and syn="bc" rendered to uniquely specify the depenedency relationships among the children (elements and words).
• Ver. 0.27 (1998-12-09)
  • </su> restored.
• Ver. 0.26 (1998-12-05)
  • Semantic relation mob (measured object) added.
• Ver. 0.25 (1998-11-22)
  • <abbr> and <abbrp> deleted.
  • `2. Global Attribute' and `5. Syntactic Constructions' merged to `2 General Attributes'.
  • Syn explicitly stated to be a gloal attribute.
  • `6. Communicative Functions' moved to Section 11.
• Ver. 0.24 (1998-11-20)
  • Delete <su> and <cs>; use <adp> in place of <cs>.
  • Delete crf; use eq instead.