Presentation is loading. Please wait.

Presentation is loading. Please wait.

Temporal information extraction: Reasoning with events based on their descriptions Leon Derczynski University of Sheffield.

Similar presentations


Presentation on theme: "Temporal information extraction: Reasoning with events based on their descriptions Leon Derczynski University of Sheffield."— Presentation transcript:

1 Temporal information extraction: Reasoning with events based on their descriptions Leon Derczynski University of Sheffield

2 Introduction  Background  Anchoring events  Reasoning about events  Representing temporal data  Evaluating annotations

3 Background  Why bother?  Temporal information affects everything described by language.  The world is in a state that changes with time.  Not all assertions made in written text are true together.  Temporal information shows which sets of data can concurrently be true.

4 Tense and temporal models  Zeno Vendler (1957) “Verbs and Times”  Hans Reichenbach (1947) “The tenses of verbs”  James Allen (1983) “Maintaining knowledge about temporal intervals”

5 Vendler  Vendler verb classification: Verb instances fall into one of four groups:  Stative: a persistent state (“John sits”)  Activity: lasts for a finite period (“Bob ran for an hour”)  Accomplishment: takes a finite period, and culminates (“Kate climbed the hill in five minutes”)  Achievement: Instantaneous finishing events (“Lucy reached the top of Everest”) Tests are provided to see which group a verb sense fits in.

6 Reichenbach  Reichenbach model of verb tenses: Speech time: when the words were uttered. Event time: when the described event occurred. Reference time: like a viewpoint. “The cat will break the door” – ST=RT, ET in the future “The cat will have broken the door” – ST = present, ET in the future, RT looks back onto ET Allows simplistic description of any phrase. Tracking reference time is sometimes very helpful: “When John comes home, I will have gone” In this case, when describes a reference time for the whole sentence.

7 Allen  Interval logic: All events are described as intervals, with start and end points. Interval relation types are defined (before, includes, starts…). A table for inferring about interval relations is given.  E.g.: A before B, A includes D: Before stipulates that A’s endpoint is before B’s start. We can infer D before B.

8 Anchoring events  Introduction to event anchoring  Dealing with named weekdays  TEA – an implemented anchoring system  Problems

9 Anchoring events  Fixing information from text to a timeline.  Calendrical time is a common reference, given a calendar.  Expressions describing a time are sometimes referred to as TIMEXs.  Once identified, a TIMEX may be normalised to a fully specified date or interval. Named entity recognition, finite state grammars and machine learning have all been used to identify these expressions.  Appropriate granularity should be chosen.

10 Weekday references  The English week has seven day names.  A single day name is often deemed sufficient reference for a human: “I’ll see you next Tuesday” “Monday, and the markets are buzzing”  To anchor a weekday, given ST and a day name, we need to choose direction from ST, and optionally distance. Baldwin: an inclusive 7-day sliding window, centred on today. Mani & Wilson: find controlling verb’s tense and use this to determine direction. Tense estimation: check PoS of sentence tokens for VBD; if found, assume backwards. Dependency-based: use Stanford parser to find controlling verb. Mazur & Dale - “What’s the Date?” (2008)

11 Generic vs. specific  Some expressions, that look like TIMEXs, should not be normalised.  “Today” can mean: the 24-hour period containing ST and bounded by midnights. Modern times, or a change of frame of reference:  “In Victorian times, ladies wore long dresses. Today, modern fashions do not dictate a single length.”  This second idea is not restricted to the period from 00:00 to 23:59 GMT on Thursday 7 th May 2009!  As 90% of uses in some texts are specific 1, some systems choose to accept a 10% error rate.  Features based on local words can help distinguish generic from specific, but below this baseline accuracy. 2 1: Han, Gates & Levin – “From Language to Time: A Temporal Expression Anchorer” (2006) 2: Mani & Wilson – “Robust Temporal Processing of News” (2000)

12 TEA  Temporal Expression Anchorer: Han, Gates & Levin at CMU.  Calendar used as time ontology, dealing with various levels of granularity.  Processes TCNL (Time Calculus for Natural Language).  Identifies temporal expressions in input, and associates TIMEXs with their textually nearest verb.  Absolute and relative expressions are evaluated using TCNL: “Friday last week” is split, into “Friday” and “last week” {fri} + {now - |1week|} = {fri,{now - |1week|}} = {now - |1fri|}  Constraint satisfaction based on a calendar model narrows the possible set of absolute dates.

13 Determining event durations  Given some normalised expressions, knowing event durations can greatly increase our reasoning ability.  Data can be taken from human annotators.  Determining a typical event duration is difficult: “The dog ran up the hill” “Linda had finished her cleaning” This results in low inter-annotator agreement.  A simplified approach would allocate durations into two classes: shorter or longer than a day.  Possible to classify events this simply with 76% accuracy, using hypernym and local word PoS features. Pan, Mulkar & Hobbs – “Learning Event Durations from Event Descriptions” (2006)

14 Reasoning about events  Introduction  Temporal closure  Minimal notations and temporal inference  Help from linguistic models

15 Reasoning about events  Annotations often only describe a subset of a document’s temporal information, perhaps as a number of labelled events and times.  An annotation may also include some links between pieces of temporal information.  It is possible to infer data about relations between points, given a set of rules or logic, and some existing relations.  It is also possible to add detail and boundaries to an annotation based on linguistic features of the source text.  This ability to reason about events saves human annotators work, and allows us to maximise the available descriptions from their efforts.

16 Temporal closure  A temporal closure can be thought of as a graph: Times and events are node; relations are edges. Every time and event is connected to every other. E.g.  t1 is Tuesday 5 th May 2009  e1 is hearing this talk We can say: t1 before e1, thus giving a type to this relation.  A temporal closure includes relations between every node in the graph.  This can lead to very large amounts of data for only a moderate-sized document.

17 Minimal annotations  It is rare for every relation (graph edge) to be annotated. We can infer some relations: (t1 before e1) ^ (e1 before e2) => (t1 before e2)  Inference can be used to complete a closure without specifying every relation’s type.  When this applies, and no more relations can be removed, we have a minimal annotation.  For example: Three nodes: e4, e5, e6 Closure has 3 possible relations A minimal graph may just say:  (e4 after e5)  (e5 simultaneous e6) To infer the closure, we simply need to add:  (e4 after e6), or (e6 before e4)

18 Relation inference  Allen’s interval logic describes 13 relationships, and provides a transitivity table for inferring a relation given two related ones. Some inconsistent labellings are possible. Backtracking over the initial graph should detect these cases.  A set of ten inference rules can be used: Allen’s 13 relations are reduced to just 3, including some reversal of parameters.  Only before, simultaneous and includes are used  e9 after e10 => e10 before e9 These rules can be iteratively added to an agenda and used to reason with a database of approved relations.  For small graphs (< 2000 edges) we can assign types to around 10% of relations, given a human annotation.

19 Applying Reichenbach  Reference time can provide a boundary on an event. “John had eaten all the pies”  Event 1 = eating  ET – RT – ST “John had eaten all the pies when Annika arrived”  Event 2 = arriving Reference time is the same across the sentence.  ET – RT = ET2 – ST Because we know that RT is after ET and equal to ET2, we can specify three temporal relations:  e1 before e2  e1 before ST  e2 before ST  Having a model for tenses allows us to confidently add relations to a temporal graph of a discourse.

20 Representing temporal data  Introduction  TIMEX and TimeML  TCNL  T-BOX

21 Representing temporal data  Once we can identify temporal information, we need to store this information.  Temporal information is rich, and favours a format that can capture it well. Aspect, polarity, tense, part of speech Event class, event frequency Hints about reference, speech and event time  Notation languages are available both for storing and working with this data.  These languages are new (under a decade old), and possibly not yet mature.

22 TIMEX  Standard for describing a time-specific expression.  Evolved through the MUC conferences and TERN, through TIMEX, TIMEX2 and TIMEX3.  TIMEX3 is currently used as the means of describing absolute times in TimeML. 10/30/89

23 TimeML  SGML-based language for temporal annotation.  Allows identification of events and times.  Thorough provision of links between events and times: TLINK: temporal, possibly including a SIGNAL tag to a linking word SLINK: subordinate ALINK: aspectual  ISO standard.

24 TimeML - TimeBank  Corpus of 181 newswire texts.  Temporal information annotated in TimeML: 6383 TLINKs, 7940 EVENTs, 3004kB in size.  Tiny compared to some other types of corpus.  Involved a large human annotator effort and a few different versions.  Biggest temporally annotated corpus.

25 TCNL  Developed at CMU with L. Levin.  Useful for reasoning between events.  Captures intensional meanings of expressions. “Yesterday” becomes {now-|1day|} instead of something like 20090506  A set of operators are used to reason between operands: +/- for forward/reverse shifting @ for in;  {|2sun| @ {may}} is “the second Sunday in May” & for distribution;  {15hour} & [{wed}:{fri}]} is “3pm from Wednesday to Friday”

26 T-BOX  Reading solid SGML is inconvenient for humans; a visual representation of events may be preferable.  Presenting events on a timeline may lead to unintentional over-specification. Suggests a distance. Many intervals are left with one end open Plotting parts of a sentence in temporal order will destroy word order, making it hard to read  Annotating documents can be done more easily when events are grouped locally and visually connected.  T-BOX 1 from Brandeis specifies a set of rules for rendering events and their relations. 1: Verhagen – “Drawing TimeML relations with T-BOX” (2007)

27 T-BOX  Relations only exist between nodes that are directly connected or contained. This suggests: - X contains Y - Y is before Z  Drawing a temporal closure could provide a very cluttered and messy graph.  A set of guidelines are provided for reducing graphs to something more visually appealing. Equivalence classes for some events. Break cycles in graphs. Remove derivable relations.

28 Evaluating annotations  Typical annotation evaluation.  Graph-based evaluation.

29 Evaluating annotations  Annotations can be compared in different ways.  When evaluating automated TIMEX or relation identification against a gold standard, we can measure precision and recall. TimeBank is often used as a gold standard for training and evaluation or systems working in TimeML.  Evaluating TIMEX normalisation needs a different measure, as there are varying degrees of correctness available.

30 Graph based evaluation  Based on the use of minimal temporal graphs.  Graphs between events (intervals) are converted into graphs between points: Smaller set of relations, needing only = and < Simpler algebra  Simultaneous points are grouped into nodes.  Graphs over the same set of points can then be compared, based on the number of node splits and merges needed to reach one from the other.

31 Summary  Background and models useful for temporal information extraction.  Technical approaches to temporal IE.  How to reason about events.  Temporal closure & minimal annotations.  Notations for temporal information.  Evaluating temporal graphs & annotations.

32 Questions


Download ppt "Temporal information extraction: Reasoning with events based on their descriptions Leon Derczynski University of Sheffield."

Similar presentations


Ads by Google