A RankRewrite Framework for XML Data Summarization

1
A Rank-Rewrite Framework for XML Data
Summarization

• Maya Ramanath
• Kondreddi Sarath Kumar
• Max-Planck Institute for Informatics
• Saarbruecken, Germany

2
Example Text Document news report
Egyptians Suffer Second Air Tragedy in a
Year CAIRO, Egypt -- The crash of a Gulf Air
flight that killed 143 people in Bahrain is a
disturbing deja vu for Egyptians It is the
second plane crash within a year to devastate
this Arab country. Sixty-three Egyptians were on
board the Airbus A320, which crashed into shallow
Persian Gulf waters Wednesday night after
circling and trying to land in Bahrain. On Oct.
31, 1999, a plane carrying 217 mostly Egyptian
passengers crashed into the Atlantic Ocean off
Massachusetts. The cause has not been determined,
providing no closure to the families, whose grief
was reopened this month with the release of a
factual report by the National Transportation
Safety Board. Walid Mourad, head of the Egyptian
Pilots Association and a voice often heard in
relation to the EgyptAir investigation, said
Wednesday's crash is a tragedy for the Arab
people as a whole. "We are all family and
brothers. We all have something in this," Mourad
said. "But for the Egyptians, this is a double
blow. Two disasters in a row for the Egyptians."
Many of the passengers on the Gulf Air flight
were headed for jobs in Bahrain or elsewhere in
the Gulf. Rida Hassan was one of those escaping
Egypt's moribund economy for work in the oil-rich
Gulf. Hassan's uncle said he rushed to the Cairo
airport after hearing a list of the passengers
read on television. The uncle, who would not give
his name, said his nephew had come home to get
married and stayed only a month. In the hours
just after the crash, relatives at the Cairo
airport expressed anger and frustration at Gulf
Air for the slow release of information. Women
screamed and men tried vainly to calm them. "No
information is being given to us. Absolutely
nothing," Mohammed Ibrahim el-Naggar said hours
after the crash. "We were told that there were
some survivors but no names were given."
El-Naggar said his cousin, her husband who works
in Dubai, and their two children aged 2 and 3
were on the downed plane. Gulf Air said it was
sending a special plane to carry 134 relatives to
Manama airport later Thursday. "All necessary
measures have been taken to receive the families
of the victims," Mohammed al-Sayed Abbas, the
Egyptian ambassador to Bahrain, told Egyptian
television. "The embassy staff will be with them
step by step until they identify the deceased."
In Bahrain, relatives were beginning the
wrenching process of identifying the victims from
photographs taken after the bodies were retrieved
from the Gulf. Egypt, which lacks the oil wealth
of the Gulf and has an economy struggling to
revive from decades of socialist stagnation, has
a long tradition of sending workers to the Gulf
to fill everything from skilled to menial jobs.
Remittances from citizens working abroad make up
Egypt's biggest source of foreign exchange.
3
Important Sentences
Egyptians Suffer Second Air Tragedy in a
Year CAIRO, Egypt -- The crash of a Gulf Air
flight that killed 143 people in Bahrain is a
disturbing deja vu for Egyptians It is the
second plane crash within a year to devastate
this Arab country. Sixty-three Egyptians were on
board the Airbus A320, which crashed into shallow
Persian Gulf waters Wednesday night after
circling and trying to land in Bahrain. On Oct.
31, 1999, a plane carrying 217 mostly Egyptian
passengers crashed into the Atlantic Ocean off
Massachusetts. The cause has not been determined,
providing no closure to the families, whose grief
was reopened this month with the release of a
factual report by the National Transportation
Safety Board. Walid Mourad, head of the Egyptian
Pilots Association and a voice often heard in
relation to the EgyptAir investigation, said
Wednesday's crash is a tragedy for the Arab
people as a whole. "We are all family and
brothers. We all have something in this," Mourad
said. "But for the Egyptians, this is a double
blow. Two disasters in a row for the Egyptians."
Many of the passengers on the Gulf Air flight
were headed for jobs in Bahrain or elsewhere in
the Gulf. Rida Hassan was one of those escaping
Egypt's moribund economy for work in the oil-rich
Gulf. Hassan's uncle said he rushed to the Cairo
airport after hearing a list of the passengers
read on television. The uncle, who would not give
his name, said his nephew had come home to get
married and stayed only a month. In the hours
just after the crash, relatives at the Cairo
airport expressed anger and frustration at Gulf
Air for the slow release of information. Women
screamed and men tried vainly to calm them. "No
information is being given to us. Absolutely
nothing," Mohammed Ibrahim el-Naggar said hours
after the crash. "We were told that there were
some survivors but no names were given."
El-Naggar said his cousin, her husband who works
in Dubai, and their two children aged 2 and 3
were on the downed plane. Gulf Air said it was
sending a special plane to carry 134 relatives to
Manama airport later Thursday. "All necessary
measures have been taken to receive the families
of the victims," Mohammed al-Sayed Abbas, the
Egyptian ambassador to Bahrain, told Egyptian
television. "The embassy staff will be with them
step by step until they identify the deceased."
In Bahrain, relatives were beginning the
wrenching process of identifying the victims from
photographs taken after the bodies were retrieved
from the Gulf. Egypt, which lacks the oil wealth
of the Gulf and has an economy struggling to
revive from decades of socialist stagnation, has
a long tradition of sending workers to the Gulf
to fill everything from skilled to menial jobs.
Remittances from citizens working abroad make up
Egypt's biggest source of foreign exchange.
4
Possible Summary Generated by MEAD, Radev et al.

• CAIRO, Egypt -- The crash of a Gulf Air flight
  that killed 143 people in Bahrain is a disturbing
  deja vu for Egyptians It is the second plane
  crash within a year to devastate this Arab
  country. Sixty-three Egyptians were on board the
  Airbus A320, which crashed into shallow Persian
  Gulf waters Wednesday night after circling and
  trying to land in Bahrain. On Oct. 31, 1999, a
  plane carrying 217 mostly Egyptian passengers
  crashed into the Atlantic Ocean off
  Massachusetts. Many of the passengers on the Gulf
  Air flight were headed for jobs in Bahrain or
  elsewhere in the Gulf. In the hours just after
  the crash, relatives at the Cairo airport
  expressed anger and frustration at Gulf Air for
  the slow release of information.

5
General Technique

• Identify text spans
• Rank the text spans
• Include them in the summary
• Rewrite if required

6
Desired Features

• Importance
• Non-redundant
• Coherent
• Coverage

7
Outline

• Introduction
• XML data summarization
• Ranking text and tags
• Generating Summary
• Evaluation
• Discussion

8
Challenges

• Features involve structure and value
• They have different characteristics
• Data-oriented vs. Document-oriented
• Text summarization techniques can be re-used, but
  are not enough
• Lots of redundancy
• Structural redundancy is inherent
• Rewriting
• Involves both structure and value

9
Example Document

• 2001 A Space Odyssey
• title, production_year, director, name, job,
  colourinfo, production_country,
  production_language, genre, plot, soundmix,
  alternative_title, alternate_version, goof,
  keyword

10
title
2001 A Space Odyssey
prod_year
1968
name
Kubrick, Stanley
producer
job
producer
director
Kubrick, Stanley
actor
Dullea, Keir
cast
casting
role
Dave Bowman
actor
Lockwood, Gary
movie
cast
casting
role
Dr. Frank Poole
actor
Sylvester, William
cast
casting
role
Dr. Haywood R. Floyd
prod_location
Shepperton Studios, Shepperton, Surrey, England,
UK
prod_location
Borehamwood, Hertfordshire, England, UK
prod_location
Isle of Harris, Scotland, UK
prod_location
Monument Valley, Utah, USA
This movie is concerned with intelligence as the
division between animal and human, then asks a
question what is the next division? Technology is
plot
11
title
2001 A Space Odyssey
prod_year
1968
name
Kubrick, Stanley
producer
job
producer
director
Kubrick, Stanley
actor
Dullea, Keir
cast
casting
role
Dave Bowman
actor
Lockwood, Gary
movie
cast
casting
role
Dr. Frank Poole
actor
Sylvester, William
cast
casting
role
Dr. Haywood R. Floyd
prod_location
Shepperton Studios, Shepperton, Surrey, England,
UK
prod_location
Borehamwood, Hertfordshire, England, UK
prod_location
Isle of Harris, Scotland, UK
prod_location
Monument Valley, Utah, USA
This movie is concerned with intelligence as the
division between animal and human, then asks a
question what is the next division? Technology is
plot
12
movie
title
movie
prod_year
movie
producer
name
movie
producer
job
movie
director
cast
casting
movie
actor
role
cast
casting
movie
prod_location
movie
plot
movie
13
2001 A Space Odyssey
1968
Kubrick, Stanley
producer
Kubrick, Stanley
Dullea, Keir
Dave Bowman
Lockwood, Gary
Dr. Frank Poole
Sylvester, William
Dr. Haywood R. Floyd
Shepperton Studios, Shepperton, Surrey, England,
UK
Borehamwood, Hertfordshire, England, UK
Isle of Harris, Scotland, UK
Monument Valley, Utah, USA
This movie is concerned with intelligence as the
division between animal and human, then asks a
question what is the next division? Technology is
14
movie
title
2001 A Space Odyssey
movie
prod_year
1968
movie
producer
name
Kubrick, Stanley
movie
producer
job
producer
movie
director
Kubrick, Stanley
Dullea, Keir Lockwood, Gary Sylvester, William
cast
casting
movie
actor
David Bowman Dr. Frank Poole Dr. Haywood Floyd
role
cast
casting
movie
Shepperton Studios, Shepperton, Surrey, England,
UK Borehamwood, Hertfordshire, England, UK Isle
of Harris, Scotland, UK Monument Valley, Utah, USA
prod_location
movie
This movie is concerned with intelligence as the
division between animal and human, then asks a
question what is the next division? Technology
is
plot
movie
15
Ranking what is it?

• Ranking based on importance
• Importance based on frequency of occurrence,
  representativeness, popularity

16
Types of Text

• Named Entities
• actor, role, director, name

Dullea, Keir Lockwood, Gary Sylvester, William
David Bowman Dr. Frank Poole Dr. Haywood Floyd

• Others (text which can be converted to terms)
• plot, production_location, trivia_item

Shepperton Studios, Shepperton, Surrey, England,
UK Borehamwood, Hertfordshire, England, UK Isle
of Harris, Scotland, UK Monument Valley, Utah, USA
17
Ranking Context
Tag context
Document context
Corpus context
18
Named Entities

• actor
• List of 50 actors (and corresponding roles)
• Goal Pick top 2
• Method choose context, compute popularity
• Context document (augmented with corpus, if
  required)
• Result actors Dullea, Keir (1)
• Lockwood, Gary (2)

lttriviagt
According to Keir Dullea (Dave Bowman),
Davenport, Nigel and Balsam, Martin were hired
and later replaced before Rain, Douglas finally
landed the role of HAL. Nigel Davenport was
actually was on-set in England during filming,
reading HALs lines off-camera so that Dullea and
Gary Lockwood could react to them.
19
Others (text converted to terms)

• production_location
• Shepperton Studios, Shepperton, Surrey, England,
  UK
• Borehamwood, Hertfordshire, England, UK
• Isle of Harris, Scotland, UK
• MGM British Studios, Borehamwood, Hertfordshire,
  England, UK
• Monument Valley, Utah, USA

uk 4 england 3 hertfordshir 2 borehamwood
2 studio 2
Goal Pick 2 samples
production_location MGM British Studios,
Borehamwood, Hertfordshire, England,
UK Borehamwood, Hertfordshire, England, UK
20
Important Tags

• Tags which are salient in the corpus

21
Outline

• Introduction
• XML data summarization
• Ranking text and tags
• Generating Summary
• Evaluation
• Discussion

22
Summary Generation

• Given A ranked list of tags and text, size
  constraint
• Required A good summary

23
Summary Generation

• Generate first-cut summary
• Maintain tag proportions as in the original
  document
• Fill in values according to rankings
• Iterative refinement
• Remove least important text unit of least
  important tag
• Repeat until required size is reached

24
Step 1 Initial Summary
title
2001 A Space Odyssey
prod_year
1968
name
Kubrick, Stanley
producer
job
producer
director
Kubrick, Stanley
actor
Dullea, Keir
cast
casting
role
Dave Bowman
actor
movie
Lockwood, Gary
cast
casting
role
Dr. Frank Poole
Shepperton Studios, Shepperton, Surrey, England,
UK
prod_location
25
Step 2 Iterative Refinement
title
2001 A Space Odyssey
prod_year
1968
name
Kubrick, Stanley
producer
job
producer
director
Kubrick, Stanley
actor
Dullea, Keir
cast
casting
actor
movie
Lockwood, Gary
cast
casting
26
Outline

• Introduction
• XML data summarization
• Ranking text and tags
• Generating Summary
• Evaluation
• Conclusion

27
Data Sets

• IMDB documents
• Titanic (200KB)
• Fear and Loathing in Las Vegas (77KB)
• Fight Club (85KB)
• 2001 A Space Odyssey (129KB)
• DBLP documents

28
Setup and Metrics

• 512B and 1KB summaries
• 3 evaluators per document
• Intrinsic Evaluation
• Grade summary quality from 1 (best) to 5 (worst)
• Top-5 evaluation
• Top-5 tags
• Top-5 values (IMDB actor)

29
Results for IMDB
30
Conclusions and Ongoing Work

• Ranking Text
• Diversity in choosing sample values
• Choosing better named entity samples
• Ranking Tag
• Special tags (oscar_winner)
• Summary Generation
• Current method is not best suited for high
  coverage
• Evaluation

31
THANKS!
32
Enumerate
Variant Find Sample, then Enumerate
genre
ltmoviegt ltgenregt Sci-Fi lt/genregt ltgenregt
Adventure lt/genregt lt/moviegt
Sci-fi
movie
genre
Adventure
ltmoviegt ltgenregt Sci-Fi Adventure
lt/genregt lt/moviegt
Sci-fi Adventure
movie
genre
33
Shorten

• ORIGINAL (plot)
• Moon explorers encounter a monolith that points
  them to a destination near Jupiter. In flashback,
  we see another size of the monolith playing a key
  role in human evolution, i.e., we learn how to
  kill. An expedition is launched to investigate
  the Jupiter possibility. Two young astronauts and
  a bunch in suspended animation spend months in
  space, passing their time partly in communicating
  with the human-like brain of their ship's
  computer, HAL. HAL malfunctions and causes the
  death of all the suspended animation passengers
  as well as one of the "awake" astronauts the
  other one barely survives and figures out how to
  disable HAL. He arrives alone at the Jupiter
  destination and undergoes a series of
  cinematographically confusing experiences that
  amount to his final appearance on the screen as a
  giant fetus.
• 50 SUMMARY
• Moon explorers encounter a monolith that points
  them to a destination near Jupiter. Two young
  astronauts and a bunch in suspended animation
  spend months in space passing their time partly
  in communicating with the human-like brain of
  their ship's computer HAL. HAL malfunctions and
  causes the death of all the suspended animation
  passengers as well as one of the awake astronauts
  the other one barely survives and figures out how
  to disable HAL.

Variants Find Samples, then Shorten (review)
Shorten, then Enumerate
(alternate_version)