Chronology of an Epidemic

1
Chronology of an Epidemic
Summary
Sources Gulu 14 Oct 2000 232501 On 13 Oct
2000, it was reported that at least 30 people in
this northern Uganda town have died in recent
weeks of a hemorrhagic fever that authorities say
16 Oct 2000 114234 GMT So far 10 people
have died in hospital, including 3 nurses
treating the sick. The other victims have
succumbed in their villages before they could get
to medical help. Nairobi 17 Oct 2000 093341
Fears of the deadly Ebola fever gripped Kenya on
Monday as 2 children died of The disease is
reported to have killed 37 people in northern
Uganda at the weekend. Nairobi 17 Oct 2000
213742 By Tue 17 Oct 2000, Radio Uganda was
reporting 73 known cases, of whom 37 had died
Geonode
2
Recap Chronology

• Narrative summaries at the level of chronology
  can be generated for arbitrary texts by a
  combination of NLP and temporal reasoning
• Inter-annotator reliability on links is an issue,
  although temporal reasoning and visualization can
  help

3
II. Story

• So far, we have seen temporal constraint networks
  where events are anchored in time
• However, a narrative is more than a temporal
  constraint network
• Every narrative has a discourse-level structure
• We have already seen that some links require
  discourse-level knowledge
• We can represent this structure formally as a
  Temporal Discourse Model Mani and Pustejovsky
  2004

4
A Temporal Discourse Model

• Edmond made his own Christmas presents this year.
  
• First he dried a bunch of tomatoes in his oven.
• Then he made a booklet of recipes that use dried
  tomatoes.
• He scanned in the recipes from his gourmet
  magazines.
• He gave these gifts to his family

constraints Ea lt Ee, Eb lt Ec
Source from Spejewski 1994 Ph.D. diss
5
A Story-Level Summary

• Summary
• at depth 1
• a. Edmond made his own Christmas presents this
  year
• e. He gave these gifts to his family.

• Source
• Edmond made his own Christmas presents this year.
  
• First he dried a bunch of tomatoes in his oven.
• Then he made a booklet of recipes that use dried
  tomatoes.
• He scanned in the recipes from his gourmet
  magazines.
• He gave these gifts to his family.

6
Summary of Childrens Story

• a. David wants to buy a Christmas present for a
  very special person, his mother.
• b. David's father gives him 5.00 a week pocket
  money and
• c. David puts 2.00 a week into his bank
  account.
• d. After three months David takes 20.00 out of
  his bank account and
• e. goes to the shopping mall.
• f. He looks and looks for a perfect gift.
• g. Suddenly he sees a beautiful brooch in the
  shape of his favorite pet.
• h. He says to himself
• i. "Mother loves jewelry, and
• j. the brooch costs only l7.00."
• k. He buys the brooch and
• l. takes it home.
• m. He wraps the present in Christmas paper and
• n. places it under the tree.
• o. He is very excited and
• p. he is looking forward to Christmas morning to
  see the joy on his mother's face.
• q. But when his mother opens the present
• r. she screams with fright because
• s. she sees a spider.

• David puts 2.00 a week into his bank account c
• to buy a Christmas present for a
• his mother a.
• After three months David takes 20.00 out of his
  bank account and d
• buys the brooch k
• But when his mother opens the present q
• she screams with fright because r
• she sees a spider.s

7
Recap Story Summarization

• Narrative summaries can be generated, in
  principle, based on a new tree-based
  representation (called Temporal Discourse Models)
  for discourse structure of narrative
• Story condensation can be based on tree depth or
  other metrics
• By focusing on temporal aspects of narrative,
  temporal relations can serve as surrogates for
  discourse relations
• This can help constrain event ordering
• TDMs can be built on top of TimeML and its
  annotation tools
• One can expect good inter-annotator reliability
  on TDMs given the transparent nature of the tree
  structure along with clear annotation conventions

8
Modern Times

• A Film by Charles Chaplin

9
Modern Times human summary and TLINKs
10
Narrative Complexity L1 versus L2 speakers

• Written narrative retelling task on Modern Times
• 16 speakers of English 8 native speakers and 8
  students in Advanced courses in an Intensive
  English Program.
• Differences between native speakers (NS) and
  non-native speakers (NNS) are significant (p
  .03) on Wds/TLink.
• On average, NSs use fewer wds to create TLinks
  (8.2/TLink vs. 10.1 for NNSs).
• A slightly higher proportion of NS Events are
  TLinked (73 vs 66 for NNSs p .05)
• NS have cleaner narrative structure
• Number of closed TLINKS for NS far exceeds the
  number for NNS (12,330 vs. 4924)
• This means NS have, on the average, longer chains
  of TLINKS
• 
  Connor-Linton and Mani 2004

11
Whats Missing?

• Modeling non-literal language
• Modeling affect
• Modeling motives
• Modeling social values
• etc.
• etc.

12
III. Plot

• Fictional Narratives have plots
• Real-life narratives , including biographies,
  histories, etc., have plots as well!

13
The Plot Unit Approach

• a-b. John was thrilled when Mary accepted his
  engagement ring.
• John experiences a mental state of wanting to
  marry Mary (E0.1), and this causes him to
  actualize his desire by offering her an
  engagement ring (E1).
• Mary experiences a mental state of wanting to
  marry John (E0.2), and this causes her to
  actualize her desire by accepting the ring (E2).
  
• the latter has also a positive affect for John,
  who is thrilled (E3).

Based on Lehnert 1981
14
Plot Units as Explanations

• Plot Units can be viewed as templates for
  answering Why questions
• Why did John offer an engagement ring to Mary?
• Why did Mary accept it?
• Why was John thrilled when she did?
• Etc.
• This requires creating such templates, and
  formally reasoning with them

15
Johns Engagement Plot

• a-b. John was thrilled when Mary accepted his
  engagement ring.
• John experiences a mental state of wanting to
  marry Mary (E0.1), and this causes him to
  actualize his desire by offering her an
  engagement ring (E1).
• Mary experiences a mental state of wanting to
  marry John (E0.2), and this causes her to
  actualize her desire by accepting the ring (E2).
  The latter has also a positive affect for John,
  who is thrilled (E3).
• c. But when he found out about her fathers
  illegal mail-order business,
• John finds out (E4-) that Marys father is a
  crook,
• d. he felt torn between his love for Mary and his
  responsibility as a police officer.
• which motivates him to be in a mental state of
  wanting to enforce the law (E0.3), which creates
  a problem (PPUP1) for John.
• e. When John finally arrested her father
• He actualizes this desire by making an arrest
  (E5), which results in a problem resolution for
  John (PUPR1), but a negative affect (E5-) for
  Mary,
• f. Mary called off their engagement.
• Mary is motivated towards a mental state of
  wanting revenge (E0.4), which she actualizes by
  calling off the engagement (E6). This
  retaliation (PUR1) on her part results in a
  positive tradeoff (PUT1) for her. However, it
  results in a negative affect state (E6-) for
  John, terminating the positive affect state from
  his engagement, leading to a loss (PPUL1) for
  him.

16
Summarizing Plot Units

• Lehnert 1981
• Build plot unit graph for narrative
• Identify top-level plot units
• Derive a plot-unit connectivity graph
• Identify salient plot units based on connectivity
• Use canned text to generate a draft summary
• Revise the draft so as to accommodate plot units
  related to the salient ones, along with various
  textual smoothing operations.

17
Johns Engagement Plot Summary

• a. John was thrilled
• b. when Mary accepted his engagement ring.
• c. But when he found out about her fathers
  illegal mail-order business,
• d. he felt torn between his love for Mary and his
  responsibility as a police officer.
• e. When John finally arrested her father,
• f. Mary called off their engagement

• Using Lehnerts method, you get a pure extract
• Because John arrested Marys father e,
• she called off their engagement f.
• Instead, using pivotal plot unit in summary,
  i.e., retaliation
• In revenge for Johns arresting Marys crooked
  fathere,
• Mary called off their engagement f.

Lehnert(1981)
Source from Lehnert(1981)
18
A Plot Summary of Childrens Story

• a. David wants to buy a Christmas present for a
  very special person, his mother.
• b. David's father gives him 5.00 a week pocket
  money and
• c. David puts 2.00 a week into his bank
  account.
• d. After three months David takes 20.00 out of
  his bank account and
• e. goes to the shopping mall.
• f. He looks and looks for a perfect gift.
• g. Suddenly he sees a beautiful brooch in the
  shape of his favorite pet.
• h. He says to himself
• i. "Mother loves jewelry, and
• j. the brooch costs only l7.00."
• k. He buys the brooch and
• l. takes it home.
• m. He wraps the present in Christmas paper and
• n. places it under the tree.
• o. He is very excited and
• p. he is looking forward to Christmas morning to
  see the joy on his mother's face.
• q. But when his mother opens the present
• r. she screams with fright because
• s. she sees a spider.

• Davids mother screams with fright r
• when she opens the Christmas presentq
• of a spider brooch that David bought k
• and now regrets giving her.

19
Issues with Plot Units

• Limited model of emotion
• Johns feeling torn between two affects isnt
  directly expressible in Lehnerts framework.
• Positive affect for John of the arrest seems
  inappropriate, given that its embedded in a
  negative situation for John
• Brittleness of plot unit analysis
• Serious knowledge acquisition problem
• for each event, identify its possible affects for
  each human participant in the event
• determine actualization, termination, or
  motivation links

20
Recap Plot Summarization

• Plots go beyond stories in providing information
  about the positive, negative or neutral affect of
  an event for an agent.
• They also provide information about additional
  events corresponding to mental states, that
  arent mentioned explicitly in the text.
• Plot units can be viewed as templates for
  answering Why questions about events
• Plots require temporal sequencing, and so require
  story level inferences
• Parsing a text in terms of plot units requires a
  degree of commonsense reasoning
• Plot condensation can use graph-based measures,
  with a proviso that pivotal plot unit(s) be
  included in output

21
Conclusion A Research Strategy

• (Feasible) To provide chronological summaries, we
  have begun to integrate temporal information
  extraction and temporal reasoning
• This is eminently feasible, but certain kinds of
  inferences are hard to carry out
• Reliability in linguistic annotation is key
• (Harder) To summarize based on the nested
  narrative structure, discourse-level modeling is
  needed
• Here temporal information can be used to provide
  initial discourse models
• these may have to be augmented
• Reliability in linguistic annotation is key
• (Hardest) To summarize plots, an explanation of a
  story is required in terms of motive and affect
• This requires acquisition of the relevant world
  knowledge
• However, MAYBE working on this problem can make
  progress in semantics (where there has been
  little progress!!) beyond what was done in the
  roll-your-own frameworks of the 1970s

22
Resources

• Chronology
• TIMEX2 taggers Mani and Wilson 2000, Jang et al.
  2004
• TIMEX2 corpora (English TDT2, Enthusiast, UN
  parallel Spanish Enthusiast, UN parallel
  Korean) TERN data
• TLINK taggers Mani et al. 2003, 2004 Schilder
  2004 Eklund 2004
• TimeBank corpus Pustejovsky et al. 2004 186
  news documents with TimeML tags

• Story (ongoing)
• Remedia corpus (remedia.com)
• 115 documents in total, grouped into four reading
  levels, and each with a small number of questions
  for reading comprehension
• Annotated with TIMEX2 tags by MITRE Ferro 2002
• Brandeis Reading Corpus
• A collection of 100 K-8 Reading Comprehension
  articles, mined from the web
• categorized by level of comprehension difficulty
• Canadian Broadcasting Corporation (cbc4kids.ca).
• current-event stories (gt 1000) aimed at an
  audience of 8-year-old to 13-year-old students
• annotated with POS and parse tree markup
• Plot (ongoing)
• Axiomatization of plot units in event calculus

23

Thank You!