Coreference in NLP

1
Coreference in NLP

• Breck Baldwin
• breck_at_alias-i.com

2
Outline

• Phenomenon
• Relevant features
• Applying features to examples
• Algorithms

3
(No Transcript)
4
(No Transcript)
5
Pronouns

• Guiliani left Bloomberg to be mayor of a city
  with a big budget problem. Its unclear how
  hell be able to handle it during his term.

6
That References

• According to John, Bob bought Sue an Integra, and
  Sue bought Fred a Legend.
• But that turned out to be a lie. (a speech act)?
• But that was false. (proposition)?
• That struck me as a funny way to describe the
  situation. (manner of description)?
• That caused Sue to become rather poor. (event)?
• That caused them both to become rather poor.
  (combination of multiple events)?

7
Across Documents

• John Smith in Doc 1 ? John Smith in Doc 2
• DB id 12123 ? John Smith in Doc 2
• What are other classes of coreference?

8
Constraints on Coreference

• Number agreement
• Johns parents like opera. John hates it/John
  hates them.
• Person and case agreement
• Nominative I, we, you, he, she, they
• Accusative me,us,you,him,her,them
• Genitive my,our,your,his,her,their
• George and Edward brought bread and cheese. They
  shared them.

9

• Gender agreement
• John has a Porsche. He/it/she is attractive.
• Syntactic constraints binding theory
• John bought himself a new Volvo. (himself
  John)?
• John bought him a new Volvo (him not John)?
• Selectional restrictions
• John left his plane in the hangar.
• He had flown it from Memphis this morning.

10

• Gender agreement
• John has a Porsche. He/it/she is attractive.
• Syntactic constraints binding theory
• John bought himself a new Volvo. (himself
  John)?
• John bought him a new Volvo (him not John)?
• Selectional restrictions
• John left his plane in the hangar.
• He had flown it from Memphis this morning.

11

• And so doesrepeated mention
• John needed a car to go to his new job. He
  decided that he wanted something sporty. Bill
  went to the dealership with him. He bought a
  Miata.
• Who bought the Miata?
• What about grammatical role preference?
• Parallel constructions
• Saturday, Mary went with Sue to the farmers
  market.
• Sally went with her to the bookstore.
• Sunday, Mary went with Sue to the mall.
• Sally told her she should get over her shopping
  obsession.

12

• Verb semantics/thematic roles
• John telephoned Bill. Hed lost the directions
  to his house.
• John criticized Bill. Hed lost the directions
  to his house.

13
He had dodged the press for 36 hours, but
yesterday the Buck House Butler came out of the
cocoon of his room at the Millennium Hotel in New
York and shoveled some morsels the way of the
panting press. First there was a brief, if
obviously self-serving, statement, and then, in
good royal tradition, a walkabout. Dapper in a
suit and colourfully striped tie, Paul Burrell
was stinging from a weekend of salacious
accusations in the British media. He wanted us to
know he had decided after his
acquittal at his theft to trial to sell his story
to the Daily Mirror because he needed the money
to stave off "financial ruination". And he was
here in America further to spill the beans to
the ABC TV network simply to tell "my side of the
story".
14

• If he wanted attention in America, he was getting
  it. His lawyer in the States, Richard Greene,
  implored us to leave alone him, his wife, Maria,
  and their two sons, Alex and Nicholas, as they
  spent three more days in Manhattan. Just as
  quickly he then invited us outside to take
  pictures and told us where else the besieged
  family would be heading Central Park, the Empire
  State Building and ground zero. The
  "blabbermouth", as The Sun doubtless doubled up
  with envy at the Mirror's coup has taken to
  calling Mr Burrell, said not a word during the
  10-minute outing to Times Square. But he and his
  wife, in pinstripe jacket and trousers, wore
  fixed smiles even as they struggled to keep their
  footing against a surging scrum of cameramen and
  reporters. Only the two boys looked resolutely
  miserable.

15
(No Transcript)
16
A Harder Case

• Tracker
• Aliases Jayson Terril Taylor, The Game
• Properties Person, Male
• Contexts http//..., http//...
• Relationships
• 34234 (Dr.Dre)?
• 21433 (Jay-Z)?
• ...

17
(No Transcript)
18
Fixing The Game

• Find common phrases in English
• The Game common
• Jay-Z is not
• If common, require supporting context via
• Proper Noun part-of-speech
• Speculative entity detection has to have found it
  as well
• Could also use related entities

19
Neuroligin 4 (NLGN4) is a member of a cell
adhesion protein family that appears to play a
role in the maturation and function of neuronal
synapses. Mutations in the X-linked NLGN4 gene
are a potential cause of autistic spectrum
disorders, and mutations have been reported in
several patients with autism, Asperger syndrome,
and mental retardation. We describe here a family
with a wide variation in neuropsychiatric illness
associated with a deletion of exons 4, 5, and 6
of NLGN4. The proband is an autistic boy with a
motor tic. His brother has Tourette syndrome and
attention deficit hyperactivity disorder. Their
mother, a carrier, has a learning disorder,
anxiety, and depression. This family demonstrates
that NLGN4 mutations can be associated with a
wide spectrum of neuropsychiatric conditions and
that carriers may be affected with milder
symptoms. NLGN4 human 57502 a
fruit fly 32103,, fruit fly 252438,,
house mouse 50518 X fruit fly
31557,, fruit fly 42032,, fruit fly
45451,, fruit fly 45552 The Norway rat
25085 an house mouse 23803
anxiety house mouse 493091 can house
mouse 12329
20
(No Transcript)
21
(No Transcript)
22
Using Features

• Old school within-document
• Semantic Pass
• Antecedent Ordering
• Pick highest match
• Hobbs, Lappin Lease
• My contributionalso Old School
• Impose a partial order on candidate antecedents
• Pick nothing in the face of ties
• High precision at decent recall possible

23
New School

• Reason over all possible coreference relations as
  sets (within-doc)?
• Culotta, Hall and McCallum '07
• Reasoning over proper probabalistic models of
  clustering (across doc)?
• Haghighi and Klein '07