Chunking

1
Chunking

2
Chunking

• is an efficient and robust method for
• identifying short phrases in text (chunks)
• Chunks are non-overlapping spans of text,
  containing
• a head word (noun, proper name, adjective, )
• adjacent modifiers and function words
  (adjective, determiner, preposition, )

3
Example

• I begin with an intuition when I read
• a sentence, I read it a chunk at a time
• I begin with an intuition when I
  read
• a sentence, I read it a chunk at a
  time
• I begin with an intuition when I
  read
• a sentence, I read it a chunk at a
  time

4
Motivation

• Locate information
• Extract Noun Phrases Indexing
• Extract NPs and Verbs Information Extraction
• Ignore information
• Aquisition of subcat-information
• gave NP, gave up NP in NP, gave NP up,
• gave NP NP, gave NP to NP

5
Definition of Chunks (Abney93)

• major head a content word not between a function
  word f and the word selected by f

• root highest node with major head as semantic
  head

• chunk maximal string containing a major head,
  dominated by root, not contained in other chunk

6
Problem empty categories

• The underlying grammar must assume
• null determiners
• Ø poor people forms a noun chunk
• and empty nouns.
• the poor Ø forms a noun chunk

7
Problem Center-Embedding

• Function word and semantic head may be separated
  by other noun chunks.
• der/den mehrere Milliarden Euro hohen
  Schäden
• in Johns house
• Base noun chunks may be ungrammatical.
• die im Alter nachlassenden Kräfte
  

8
Problems with Coordination

• by Abneys definition
• 
  
  
  
  
• 
  
  

• 
  
  
  
  
• if chunks may be
  multi-headed
• 
  
  

• 
  
  
  
  
• 
  
  
• if conjunctions
  are excluded from
  chunks

9
System Overview
10
Base Nouns Chunks vs. Full Noun Chunks

• Base Noun Chunk maximal string dominated by root
  containing noun as major head ( Abney 1993)
• Full Noun Chunk Part of NP between determiner
  and (first) head noun (Schmid and Schulte im
  Walde 2000)
• includes names
• the discoverer Christopher Columbus
• but not coordinated NPs
• parts of Scotland and Northern Ireland
• and not appositions
• Christopher Columbus, the famous discoverer,

11
Negative Definition of Full Noun Chunks

• NP/PP stripped of - adverbials at the front and
  - PPs and relative clauses at the back
  (Brants,
  1999)
• coordinations and appositions
• ? parts ? of Scotland and Northern Ireland
• pre- and postnominal genitives
• Marias Version der Geschichte
  
• measure phrases 20 Dollar Strafe
  
  

12
Disambiguation in the recognition of full noun
chunks

• POS ambiguities resolved by POS tagger.
• PP attachment ambiguities are kept underspecified
  (wrt. positive and negative definition of full
  noun chunk).
• All other ambiguities are resolved using the
  longest-match criterion (Abney, 1993).
• Chunks should be as long as possible.
• But approaches to deal with them explicitly, by
  underspecification, or by non-monotoniciy exist.

13
Recognizing Full Noun Chunksexplicit
representation of ambiguities

• used in previous work on full noun chunking
  (Brants99, Schmid and Schulte im Walde00,
  Kermes and Evert02)
• drawback requires search
• Parser is not deterministic any longer.
• Linear complexity is lost.

14
Recognizing Full Noun Chunksdealing with
ambiguities by non-monotonic cascades

• A method retaining determinism and linear
  complexity (Schiehlen02)
• recognize base noun chunks that could form
• beginning,
• middle or
• end of a full noun chunk
• discard those noun chunks (monotonicity lost!)
• re-apply original noun chunk transducer

15
Recognizing Recursive NPsby Non-Monotonic
Cascades
16
Recognizing Recursive NPsby Underspecification

• 0 1 2 3 4
  5
• die Ende der Woche geplanten Treffen
• Underspecified Representation (à la
  Spranger05)
• lt NP, 0,1,2,3,4,5, 1,3,5, 1,3,5 gt
• Desambiguierungsalgorithmus
• Problem needs preprocessing of set of NPs

17
Case Checking
18
3 Approaches to Agreement Checking in FS Parsers

• add agreement info to POS tags and compile the
  grammar out (drawback explosion of trans table)
• postpone agreement check until after chunk
  recognition (Abney, 1997)
• interleave agreement checking with chunking
  (Neumann et al., 2000), problems with
  subcategorizing multi-words
• um Gottes willen (for God's sake)
• um takes acc., um-willen takes gen.!

19
Online Agreement Checking

• errors avoided
• genitives (case mismatch)
• in John's house
• conjunction attachment (case mismatch)
• das Leben von Schauspielern und Zirkusleuten
  the life(nomacc) of actors
  and circus people(dat)
• adjacent NPs (adjective declination)
• diese beiden ähnliche Erfolge
  
• those two(weak) similar(strong) successes

20
Online Agreement Checking

• Some grammar errors become visible only with
  agreement checking.
• N coordination is missing.
• die nachlassenden Kräfte
  the
  diminishing strength
• die Verletzungen und nachlassenden Kräfte
  the injuries
  and diminishing strength

no noun chunk!
21
Experiment by Schielen02

• Writing a finite-state grammar is worth the
  effort. FS method performs better than
  statistical method
• Noun chunker is not very good at determining POS
  tags.
• Online agreement checking improves performance.
• Shortest match is better than longest match for
  conjunction attachment.

22
Case checking Readings
23
Underspecification Context Variables

• Udo/0 1aNPnom,1bNPakk
• kennt/1
• eine/2
• nette/4
• Frau/5 1aNPakk,1bNPnom
• aus/6
• Rio/7 ADJ 1A5
• ADJ 1A1

24
Literatur

• Schiehlen02. Experiments in German Noun
  Chunking. COLING 2002, Taipei, August 27th, 2002
• Schiehlen03. A Cascaded Finite-State Parser for
  German. EACL 2003, Budapest, April 17th, 2003
• Spranger05. Combining deterministic processing
  with ambiguity awareness the case of German
  quantifying noun groups. PhD Thesis. University
  of Stuttgart.