Genia TreeBank Beta version

Scheme

1. XML format is used. For non-null elements, the syntactic category is used as tags.
   <NP>a <ADJP>very sweet </ADJP>apple </NP> Null elements are marked by using childless elements whose tags are corresponding categories. e.g. <NP NULL="NONE" ref="i10"/> Attribute NULL specifies the type of a null element.
   Value: NONE for *, Zero for 0, QSTN for *?*, T for *T*
   Null complementizers are marked with <COMP NULL="ZERO"/>
2. All labels except NX, NAC are used. The phrases that should be labelled as NX and NAC are also labeled NP.

   e.g.

   PTB: (NP:tasty (NX:(NX:California oranges) and (NX:Fuji apples)))
   -->(NP:tasty (NP:(NP:California oranges) and (NP:Fuji apples)))

3. Some function tags in PTB are not used. Those used are:
   • Tags that identify the case elements: SBJ, PRD, DTV
   • Adverbial: only TMP tag is used but others are not.
   • Other tags: CLF (�git is �c that�c �g sentence) , HLN (headline, for the UI label of Medline articles) , VOC (vocative). TPC (topicalized noun element associated with a *T*)
4. Case elements such as SBJ, PRD, DTV are attached within the tag, connected to the category name with "-".
5. Othe function tags are marked as follows.

   Adverbial

   Attribute name: SEM, Attribute value: TMP �¨SEM="TMP"
   e.g. <NP-SBJ>We </NP-SBJ>

   Others

   Attribute name: FCTN , Attribute value:

   It is..that..sentense	CLF
   Headline	HLN
   Vocative	VOC
   Vopicalized elements	TPC

6. In the gapped structures, the complete clause corresponding to the gap is used as a template, and assigned an id number (id attribute). The corresponding gap is labeled with "GAP" attribute (the value is the id number of the corresponding template. e.g.

   <S><LST>AB</LST> - <NP-SBJ id="i77"><NP>The production </NP><PP>of <NP><NP><NP>human immunodeficiency virus type 1 </NP><PRN>(<NP>HIV-1</NP>) </PRN></NP>progeny </NP></PP></NP-SBJ><VP>was <VP SYN="COOD"><VP>followed <NP NULL="NONE" ref="i77"/><PP id="i104">in <NP>the U937 promonocytic cell line </NP></PP><PP SEM="TMP">after <NP><NP>stimulation </NP><PP>either with <NP SYN="COOD"><NP>retinoic acid </NP>or <NP>PMA</NP></NP></PP></NP></PP></VP>, and <VP><PP GAP="i104">in <NP>purified human <NP SYN="COOD"><NP>monocytes </NP>and <NP>macrophages</NP></NP></NP></PP></VP></VP></VP>.</S>

7. Coordination is labeled with SYN attribute, whose value is "COOD".
8. Reference id is marked with id numbers. The refenent has the "id" attribute and the referer has the "ref" attribute with identical value with the referent's "id" attribute.
9. When the annotator is unsure bracketing tag or finds error in the original text, the comment is inserted using following attributes.
   UNSURE : when not sure about process

   attribute name:	UNSURE
   attribute value:	unsure about the name of elements:	Element_Name
   	unsure about the range of elements:	Range
   	unsure about the case elements as SBJ:	Case
   	unsure about bracketing:	Internal_Structure
   	unsure about the function tags:	Function
   	unsure about the null elements:	NullType
   	unsure about the coindex to the null elements:	Coindex
   	unsure about the null elements process:	Null
   	in case the original text is not understandable:	Over_All
   	others:	Other

   • TXTERR : when there�fs error in the original text

     attribute name:	TXTERR
     attribute value:	spelling error:	Spell
     	grammatical error:	Grammar
     	two sentences not divided:	Should_Be_Divided
     	one sentence not concatenated:	Should_Be_Concatenated
     	others:	Other

   • Other comment

     attribute name:	COMMENT
     attribute value:	free text in English

Notes

Coordination, "COOD" attribute

1. "COOD" is attached at the parent element which show the whole part of Coordination.
   PTB: (NP:(NP:a cat) and (NP:three dogs)) �¨ (NP-COOD:(NP:a cat) and (NP:three dogs))

2. When there�fs phrase coordination in object or adverbial phrase, the whole part of coordination is bracketed by "COOD" tag attached with parent element to make it clear. examples:
   
   I buy and quickly eat oranges.
   PTB: I (VP:(VP:buy (NP *RNR*-1)) and (VP:quickly eat (NP *RNR*-1)) (NP-1:oranges)). �¨I (VP:(VP-COOD:(VP:buy (NP *RNR*-1)) and (VP:quickly eat (NP *RNR*-1))) (NP-1:oranges)).
   I went out and did shopping yesterday
   PTB: I (VP:(VP:went out) and (VP:did shopping) (NP-TMP:yesterday)) �¨I (VP:(VP-COOD:(VP:went out) and (VP:did shopping)) (NP-TMP:yesterday))
3. When single-word elements of the same syntactic category are coordinated, the coodination is not explicitly marked in the original PTB. that be bracketed with flat structure. But here, single-word elementswe are also tagged "COOD".examples:
   
   PTB: (NP:John and Mary) �¨ (NP-COOD:(NP:John) and (NP:Mary))
   PTB: She (VP:smiled and then cried). �¨She (VP-COOD:(VP:smiled) and then (VP:cried)).
   PTB: I (VP:buy and eat (NP:oranges)). �¨I (VP:(VP-COOD:(VP:buy(NP *RNR*-1)) and (VP:eat(NP *RNR*-1))) (NP-1:oranges)).
   PTB: (NP:a (ADJP:beautiful, smart) cat) �¨ (NP:a (ADJP-COOD:(ADJP:beautiful), (ADJP:smart)) cat)

Noun Phrase

1. NAC and NX are not used.
2. When noun phrase consists with sequence of nouns, the internal structure is not neccessarily shown.