temporal reasoning intro to timeml cs112 october, 2004
TRANSCRIPT
Temporal ReasoningIntro to TimeML
cs112
October, 2004
TimeML: what it is
• Standard language for the mark-up of:– temporal expressions– events– temporal anchoring of events
(relations between events and temporal expressions)– temporal ordering of events
(relations between events and other events)
TimeML: goals
• Long term aim: provide the basic background for:– Temporal inference– QAS to be able to answer questions like:
• Is Gates currently CEO of Microsoft?• Were there any meetings between the hijackers and Iraq
before the WTC event?
– …
• Creation of a gold standard corpus with temporal expressions, events and basic temporal relations marked up.
An example
“Two Russians and a Frenchman left the Mir and endured a rough landing on the snow-covered plains of Central Asia on Thursday. The two Russians arrived on the Mir last August. Solovyou celebrated his 50th birthday during his six-month space voyage.”
An example
“Two Russians and a Frenchman left the Mir and endured a rough landing on the snow-covered plains of Central Asia on Thursday. The two Russians arrived on the Mir last August. Solovyou celebrated his 50th birthday during his six-month space voyage.”
What to annotate
• Time Expressions (timex)• Events• Signals• Links
We’ll mark them up with a set of attributes.
What to annotate: TIMEX3
What to annotate: TIMEX3
• Durations:– 4 hours, the whole week, half a year, …
• Calendar dates: (=points in time equal or bigger than a day)– Precise dates: March 16, 2003; two years ago today; yesterday;
– Vague dates: few days ago; ending of March;
– Week references: the 2nd week of January; several weeks later;
– Yearly quarters and halves: the 4th quarter; beginning last semester;
– Seasons: last Summer; Fall 1998;
– Year references: the 60s, 1920
– Decades, centuries and millenia: the last decade
• Times of Day: (=points in time smaller than a day)– Precise times: 7:30am; Tuesday, March 25 at 12:00pm; …
– Vague times: several minutes before
– Parts of day: Saturday afternoon, yesterday early in the morning
• Sets (reoccurring time expressions):– Two times a week, every day, …
Attributes for the tag TIMEX3
1. Timex ID (automatically assigned)
2. Type:– DATE, for expressions describing a calendar date:
the second of December, yesterday, the summer of 1971, Tuesday, ...
– TIME, for expressions describing a times of day:five minuts past eight; 7:30am; 9:00 am Friday, October 1, 1755;
– DURATION:2 months, 48 hours, all last night, three weeks.
– SET
3. Value: • ISO value for the time expression.
Feb 27, 1998 08:14 “1998-02-27T08:14:00”twelve weeks “P12W”
Attributes for the tag TIMEX3 (2)
4. Mod:
5. temporalFunction: binary attributeF: the timex provides all the info:
• Twelve o’clock January 3, 1969• Summer of 1964
T: the timex doesn’t contain all the info needed to locate it at a specific point of time (therefore, a temporal function needs to be applied):• eleven in the morning• yesterday• next year
6. anchorTimeID:– Use only when ‘temporalFunction’ attribute is set to TRUE– Refers to the temporal anchor for the “incomplete” timex.
Attributes for the tag TIMEX3 (3)
7. functionInDocument: – CREATION TIME– MODIFICATION TIME– PUBLICATION TIME– RELEASE TIME– RECEPTION TIME– NONE
• At most, once per document• Generally ‘CREATION TIME’
Attributes for the tag TIMEX3 (4)
Example of annotated TIMEXs
Mary left on Thursday and John arrived the day after.
Mary left on<TIMEX3 tid="t1" type="DATE” value="1998-WXX-4"
temporalFunction="true” anchorTimeID="t0">Thursday </TIMEX3>and John arrived <TIMEX3 tid="t2" type="DATE” value="1998-WXX-5"
temporalFunction="true” anchorTimeID="t1">the day</TIMEX3>after
TERQAS Workshop will resume Monday, July 15. The session will start at 9:00 a.m.
TERQAS Workshop will resume<TIMEX3 tid="t1" type="DATE" value="2002-07-15”temporalFunction="true" anchorTimeID="t0">Monday, July 15</TIMEX3>. The session will start at<TIMEX3 tid="t2" type="TIME" value=" 2002-07-15
T9:00" temporalFunction="true" anchorTimeID="t1">
9:00 a.m.</TIMEX3>
Example of annotated TIMEXs
What to annotate: EVENTs
What to annotate: EVENTs• Tensed Verbs:
A fresh flow of lava, gas and debris erupted there Saturday.
• Untensed verbs:Prime Minister Benjamin Netanyahu called the prime minister of the Netherlands to thank him for thousands of gas masks his country has already contributed.
• Nominalizations:Israel will ask the US to delay a military strike against Iraq until the Jewish state is fully prepared for a possible Iraqi attack.
• Adjectives:A Philippine volcano, dormant for six centuries, began exploding with searing gases, thick ash and deadly debris.
• Prepositional phrases:All 75 people on board the Aeroflot Airbus died.
• Predicative Clauses: "There is no reason why we would not be prepared," Mordechai told the Yediot Ahronot daily.
Attributes for the tag EVENT
1. Event ID: (automatically assigned)
2. Class:– REPORTING: say, report, announce, …– PERCEPTION: see, hear, watch, feel.– ASPECTUAL: begin, start, finish, stop, continue.– I_ACTION: attempt, try, promise, offer, regret,…– I_STATE: believe, want, wish,…– STATE: be on board, kidnapped, recovering, love, ..– OCCURRENCE: die, crash, build, merge, sell, take advantage of, ..
Example of annotated EVENTs
Israel may ask the United States to delay a military strike against Iraq until the Jewish state is fully prepared for a possible Iraqi attack.
Israel may
<EVENT eid="e1" class="I_ACTION”>
ask </EVENT>
the United States not to
<EVENT eid="e2" class="I_ACTION”>
delay </EVENT>
a military
<EVENT eid="e3" class="OCCURRENCE”>
strike </EVENT>
against Iraq until the Jewish state is fully
<EVENT eid="e4" class="I_STATE”>
prepared </EVENT>
for a possible Iraqi
<EVENT eid="e5" class="OCCURRENCE”>
attack </EVENT>
What to annotate: MAKEINSTANCEs
Attributes for MAKEINSTANCE
1. Event Instance ID: (automatically assigned)
2. Event ID: (automatically assigned)
3. Tense: (only pertinent for verbs)– PAST: John decided/had decided not to go.– PRESENT: John decides/has decided not to go.– FUTURE: John will decide/will have decided not to go.– NONE: John decided not to go / John’s
decision ...
4. Aspect: (only pertinent for verbs)– PROGRESSIVE: [BE + -ing] John is/was/will be eating
– PERFECTIVE: [HAVE + Ppart] John has/had/will have eaten
– PERFECTIVE_PROGRESSIVE: John has/had been eating
– NONE: John eats/ate/will eat
5. nf_morph:– NOUN: John’s decision was not to go.
– ADJECTIVE: Sue was ashamed of John’s decision.
– INFINITIVE: John decided not to go.– PRESPART: While playing, John broke his leg.
– PASTPART: Steel plates found at the scene are being analyzed by experts.
6. Polarity:– POS: no negation present John went.– NEG: negation present John did not go.
7. Modality• Modal Auxiliars like: could, can, might, may, must, should,…
8. Signal_ID --see the guidelines
9. Cardinality --see the guidelines
Attributes for MAKEINSTANCE
Example of annotated MIs
Israel may ask the United States to delay a military strike against Iraq…
Israel may
<EVENT eid="e1" class="I_ACTION”> ask </EVENT>
<MI eiid=“ei1” eid=“e1” tense=“NONE” aspect=“NONE” nf_morph=“NONE” modality=“may”/>
the United States to
<EVENT eid="e2" class="I_ACTION”> delay </EVENT>
<MI eiid=“ei2” eid=“e2” tense=“NONE” aspect=“NONE” nf_morph=“INFINITIVE”/>
a military
<EVENT eid="e3" class="OCCURRENCE"> strike </EVENT>
<MI eiid=“ei3” eid=“e3” tense=“NONE” aspect=“NONE” nf_morph=“NOUN”/>
What to annotate: SIGNALs
What to annotate: SIGNALs
• Temporal prepositions:on, in, at, from, to, before, after, during,
etc.
• Temporal conjunctions: before, after, while, when, etc.
• Temporal modifiers: (only when modifying events)
twice, every, three times, etc.
• Special characters: - and /, in temporal expressions denoting
ranges (September, 4-6, Apr. 1999/Jul. 1999, etc.).
Attributes for the tag SIGNAL
1. Signal ID: automatically assigned
(they are very boring!)
Example of annotated SIGNALs
• Temporal Prepositions, Conjunctions and Modifiers:
John taught <SIGNAL sid="s1">on</SIGNAL> Monday
All passengers died <SIGNAL sid="s1">when</SIGNAL> the plane crashed into the mountain.
What to annotate: LINKs
What to annotate: LINKs• Temporal: TLINK
It represents the temporal relationship holding between events or between an event and a timex:
Mary arrived in Boston last Thursday.
• Aspectual: ALINKIt represent the relationship between an aspectual event and its argument event.
She finished assembling the table.
• Subordination: SLINKIt is used for contexts introducing relations between an I-ACTION/I-STATE event and its event argument, or an event and a negation or modal :
She tried to buy some wine.
relType: the attribute for TLINK• Simultaneous: (only for pairs of events)
Mary was watching TV while John was frying the eggs.
• Before/After:Mary had decided not to help him. (wrt ‘watching’ or ‘frying’ in the previous sentence).
• Immediately before/Immediately after:One of the eggs crashed as soon as it touched the pan.
• Including/Being included: Mary arrived in Boston last Thursday.
• During: (states or events that persist through a duration)John fried eggs for 20 minutes on Monday.
• Beginning/Begun by:John fried eggs from 6:00pm to 6:20pm.
• Ending/Ended by:John fried eggs from 6:00pm to 6:20pm.
• Identity: (only for pairs of events)Mary was resting for a while. After her rest, she took a bath.
relType: the attribute for SLINK
• Modal, introduced mainly by:– I_States : Mary wanted John to buy some wine.– I_Actions: John tried to get ticket for the final.
• Factive, for I-events introducing some presupposition:John forgot he was in Boston last
year.
• Counter-factive, for I-events introducing a presupposition about the non-veracity of its argument:– forget (to), prevent, cancel, avoid, decline, etc.
• Evidential, mainly introduced by Reporting and Perception events:– say, report, see, hear, ..
• Negative Evidential, introduced by Reporting events that convey negative polarity:
John denied he bought only beer.
relType: the attribute for ALINK
• Initiation:John started to read.
• Culmination:John finished assembling the table.
• Termination:John stopped talking.
• Continuation:John kept talking.
Final remark
Please, read the Annotation Guidelines!