Stand-off Annotation

1
Stand-off Annotation

• Further details and examples
• Durusau and ODonnells (2001) powerpoint
  presentation
• Thompson and McKelvies (1997) Hyperlink
  semantics for standoff markup of read-only
  documents. In Proceedings of SGML Europe '97

2
Corpus basics

• Required structured, consistent, correct
  information for supporting empirical questions /
  hypothesis formation and hypothesis testing

(annotated)
corpus
collection
raw texts
source annotation
3
A frequent problem for sophisticated annotation
intersecting hierarchies

• These are some sentences with different kinds of
  things going on in them. This one is not very
  interesting.

ltSgtThese areltBgtsomeltREDgtsentenceslt/Bgt...
lt/SgtltIgtThislt/REDgtone islt/Igtnot very
interesting.lt/Sgt
4
A frequent problem for sophisticated annotation
intersecting hierarchies

• These are some sentences with different kinds of
  things going on in them. This one is not very
  interesting.

ltSgtThese areltBgtsomeltREDgtsentenceslt/Bgt...
lt/SgtltIgtThislt/REDgtone islt/Igtnot very
interesting.lt/Sgt
5
A frequent problem for sophisticated annotation
intersecting hierarchies

• These are some sentences with different kinds of
  things going on in them. This one is not very
  interesting.

ltSgtThese areltBgtsomeltREDgtsentenceslt/Bgt...
lt/SgtltIgtThislt/REDgtone islt/Igtnot very
interesting.lt/Sgt
6
Intersecting hierarchies
S
ltSgtThese areltBgtsomeltREDgtsentenceslt/Bgt...
lt/Sgt...lt/REDgt
R
B
7
A more complex example (from Durusau and
ODonnell, 2001)

• Four separate (overlapping) hierarchies

8
A more complex example

• Four separate (overlapping) hierarchies

9
A more complex example

• . Page view

1
2
10
A more complex example

• . Text view

1
2
11
A more complex example

• . Linguistic view

1
2
12
TEI Corpus Encoding Standard approach

• layered or stand-off annotation

lts from1 to13/gt ...
ltw n1gtTheselt/wgt ltw n2gtarelt/wgt ltw n3gtsomeltwgt ltw
n4gtsentenceslt/wgt ltw n5gtwithlt/wgt ...
ltb from3 to4/gt ...
ltred from4 to14/gt ...
13
A more complex example (contd)

• . Textual variant view

1
2
14
Example taken from Dialogue Work (Edinburgh
Human Language Technology Group)
Dialogue Games
Game instruct
Dialogue Moves
M instruct
M ack
M instruct
M ack
M align
M align
three
centimetres
okay
three
or
four
centimetres
okay
S1
turn
right
for
Words
right
right
S2
reparandum
repair
Disfluencies
Disfluency
15
Requirements

• To understand this all, we need
• Ways for XML markup to refer to pieces of other
  XML documents
• XPath
• XLink
• XPointer