semantics and time in language

38
Semantics and Time in Language MAS.S60 Rob Speer Catherine Havasi Some slides: James Pustejovsky

Upload: airlia

Post on 24-Feb-2016

54 views

Category:

Documents


0 download

DESCRIPTION

Semantics and Time in Language. MAS.S60 Rob Speer Catherine Havasi Some slides: James Pustejovsky. Lexical semantics. We’ve been trying to make word meanings into a functional programming language Applying functions to each other, up the parse tree, gives us a logic expression in the end - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Semantics and Time in Language

Semantics and Time in Language

MAS.S60Rob Speer

Catherine HavasiSome slides: James Pustejovsky

Page 2: Semantics and Time in Language

Lexical semantics• We’ve been trying to make word meanings

into a functional programming language• Applying functions to each other, up the parse

tree, gives us a logic expression in the end• But how do we figure out crazy functions like:

\X \y. X(\x. chase(y, x))

Page 3: Semantics and Time in Language

Being an un-parser• Work backwards from the result you want• Un-parse your way down the parse tree

Page 4: Semantics and Time in Language

“A dog barks.”• A dog barks.

exists x. (dog(x) & bark(x))

Page 5: Semantics and Time in Language

“A dog barks.”• A dog barks.

exists x. (dog(x) & bark(x))• (A dog) (barks)

A dog:barks:

Page 6: Semantics and Time in Language

“A dog barks.”• A dog barks.

exists x. (dog(x) & bark(x))• (A dog) (barks)

A dog: \P. exists x. (dog(x) & P(x))barks: \z. bark(z)

Page 7: Semantics and Time in Language

“A dog barks.”• A dog barks.

exists x. (dog(x) & bark(x))• (A dog) (barks)

A dog: \P. exists x. (dog(x) & P(x))barks: \z. bark(z)

• (A(dog)) (barks)A: dog:

Page 8: Semantics and Time in Language

“A dog barks.”• A dog barks.

exists x. (dog(x) & bark(x))• (A dog) (barks)

A dog: \P. exists x. (dog(x) & P(x))barks: \z. bark(z)

• (A(dog)) (barks)A: \Q. \P. exists x. (Q(x) & P(x))dog: \z. dog(z)

Page 9: Semantics and Time in Language

Lexical items we learnedA: \Q. \P. exists x. (Q(x) & P(x))dog: \z. dog(z)barks: \z. bark(z)

Page 10: Semantics and Time in Language

“Angus chases a dog.”Angus chases a dog: exists x. (dog(x) & chase(Angus, x))

Page 11: Semantics and Time in Language

“Angus chases a dog.”Angus chases a dog: exists x. (dog(x) & chase(Angus, x))

• Angus(chases a dog)chases a dog: \y. exists x. (dog(x) & chase(y, x)a dog: \P. exists x. (dog(x) & P(x)) from earlier slides

Page 12: Semantics and Time in Language

“Angus chases a dog.”Angus chases a dog: exists x. (dog(x) & chase(Angus, x))

• Angus(chases a dog)chases a dog: \y. exists x. (dog(x) & chase(y, x)a dog: \P. exists x. (dog(x) & P(x)) from earlier slides

• (chases) (a dog)• Let’s try to make something like this:

(\P. exists x. (dog(x) & P(x)) (\z. chase(y, z))

Page 13: Semantics and Time in Language

“Angus chases a dog.”Angus chases a dog: exists x. (dog(x) & chase(Angus, x))

• Angus(chases a dog)chases a dog: \y. exists x. (dog(x) & chase(y, x)a dog: \P. exists x. (dog(x) & P(x)) from earlier slides

• (chases) (a dog)• Let’s try to make something like this:

(\P. exists x. (dog(x) & P(x)) (\z. chase(y, z))

Page 14: Semantics and Time in Language

“Angus chases a dog.”Angus chases a dog: exists x. (dog(x) & chase(Angus, x))

• Angus(chases a dog)chases a dog: \y. exists x. (dog(x) & chase(y, x)a dog: \P. exists x. (dog(x) & P(x)) from earlier slides

• (chases) (a dog)• Let’s try to make something like this:

(\P. exists x. (dog(x) & P(x)) (\z. chase(y, z))

chases: \y. doSomethingWith(\z. chase(y, z))

Page 15: Semantics and Time in Language

“Angus chases a dog.”Angus chases a dog: exists x. (dog(x) & chase(Angus, x))

• Angus(chases a dog)chases a dog: \y. exists x. (dog(x) & chase(y, x)a dog: \P. exists x. (dog(x) & P(x)) from earlier slides

• (chases) (a dog)• Let’s try to make something like this:

(\P. exists x. (dog(x) & P(x)) (\z. chase(y, z))

chases: \X. \y. X(\z. chase(y, z))

Page 16: Semantics and Time in Language

Your turn• We add a feature grammar rule that allows for

ditransitive (two-object) verbs:VP[SEM=<?v(?obj,?pp)>] -> DTV[SEM=?v] NP[SEM=?obj] PP[+TO,SEM=?pp]

• What are the semantics of a DTV?

Page 17: Semantics and Time in Language

High-level overview of C&C• Parses using a Combinatorial Categorial

Grammar (CCG)– fancier than a CFG– includes multiple kinds of “slash rules” for gaps

and fillers– lots of grad student time spent transforming

Treebank

Page 18: Semantics and Time in Language

High-level overview of C&C• MaxEnt “supertagger” tags each word with a

semantic value• Possible semantic values for verbs determined

by VerbNet

Page 19: Semantics and Time in Language

High-level overview of C&C• Combine the resulting semantic “tags”• Find the highest-probability result with

coherent semantics• Doesn’t this create billions of parses that need

to be checked?

Page 20: Semantics and Time in Language

High-level overview of C&C• Find the highest-probability result with

coherent semantics• Doesn’t this create millions of parses that

need to be checked?• Yes. A typical sentence uses 25 GB of RAM to

find the best parse.• That’s where the Beowulf cluster comes in.

Page 21: Semantics and Time in Language

Questions about time?• The Pierre Vinken example• Events in FrameNet• Question answering

Page 22: Semantics and Time in Language

Time in Q&A• When are finals this semester?• Who is currently president of the United

States?• How many different airports has Pittsburgh

had?• How many classes have we had since January?• When did the Berlin wall fall?

Page 23: Semantics and Time in Language

Difficulties• More than 66% of times in documents are

relative• Only 15% of documents refer to the “date of

creation” (DOC)• 42% percent of the uses of the word “today”

are non-specific

Page 24: Semantics and Time in Language

James Allen• Created a temporal logic• 13 basic relations– 6 types, their inverses and equal

Page 25: Semantics and Time in Language

Allen’s Relations

Page 26: Semantics and Time in Language

Types of Information• Properties – Hold over an interval and all subintervals– “Rob was asleep all morning.”

• Events– Hold over a interval and no sub events– “Lance wrote a program last night.”

• Processes– Hold over some sub intervals– “Brett demoed during sponsor week.”

Page 27: Semantics and Time in Language

What is TimeML?• (ISO) 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)

Page 28: Semantics and Time in Language

Labeling What?• Events are taken to be situations that occur or

happen, punctual or lasting for a period of time.

• Times may be either points, intervals, or durations.

• Relations can hold between events and events and times.

Page 29: Semantics and Time in Language

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.”

Page 30: Semantics and Time in Language

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.”

Page 31: Semantics and Time in Language

Events and Relations• Event expressions; – tensed verbs; has left, was captured, will resign; – stative adjectives; sunken, stalled, on board; – Nominals: merger, Military Operation, Gulf War;

• Dependencies between events and times:– Anchoring; John left on Monday.– Orderings; The party happened after midnight.– Embedding; John said Mary left.

Page 32: Semantics and Time in Language

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.

Page 33: Semantics and Time in Language

TARSQI• Add and tag time expressions in text• TempEx (MITRE)– Determines extents and nomalizations

• GUTime (Brandeis)– Ground things like “last week”

• Evita (Brandeis)– Recognize events in time

Page 34: Semantics and Time in Language

TARSQI • GUTenLink (Georgetown)– Temporal Tagger

• Slinket (Brandeis)– Event logging

• SputLink – Based on James Allen’s time logic

Page 35: Semantics and Time in Language

Open a Document

Page 36: Semantics and Time in Language

Processed Document

Page 37: Semantics and Time in Language

Results

Page 38: Semantics and Time in Language

Making a timeline