Linguists get the abstraction, machines get the details Hal Daum - - PowerPoint PPT Presentation

▶

Oct 05, 2022 6 likes •86 views

Hal Daum III (me@hal3.name) Linguists get the abstraction, machines get the details Hal Daum III Computer Science / Linguistics University of Maryland, College Park me@hal3.name Symbol Pushing Slide 1 Hal Daum III (me@hal3.name)

SLIDE 1

Symbol Pushing Slide 1 Hal Daumé III (me@hal3.name)

Linguists get the abstraction, machines get the details

Hal Daumé III

Computer Science / Linguistics University of Maryland, College Park me@hal3.name

SLIDE 2

Symbol Pushing Slide 2 Hal Daumé III (me@hal3.name)

NLP's use of linguists, a caricature

➢

Linguists develop theory

➢

Linguists richly annotate data (eg treebank)

➢

NLP people train systems (eg parser)

➢

Parser output fed into machine translation system

➢

Machine translation system has no idea what the input symbols mean

➢

NP, VP, VBD, .... might as well be X1, X2, X3, …

SLIDE 3

Symbol Pushing Slide 3 Hal Daumé III (me@hal3.name)

Where does this model work?

➢

Works when entire pipeline is learned from data

➢

And we make no use of prior knowledge

➢

Pretty much any other time

Where does this model not work?

SLIDE 4

Symbol Pushing Slide 4 Hal Daumé III (me@hal3.name)

Inferring Tags from the Structure

➢ INPUT: ➢ OUTPUT: ➢ Baseline: ➢ Random guessing: 4% accuracy

The man ate a big sandwich D

N V D J N

SLIDE 5

Symbol Pushing Slide 5 Hal Daumé III (me@hal3.name)

Sources of Knowledge

➢ Seeds (frequent words for each tag)

➢ N: membro, milhoes, obras ➢ D: as [the,2f] o [the,1m] os [the,2m] ➢ V: afector, gasta, juntar ➢ P: com, como, de, em

➢ Typological rules:

➢ Art ← Noun ➢ Prp → Noun

➢ Tag knowledge:

➢ Open class ➢ Closed class

SLIDE 6

Symbol Pushing Slide 6 Hal Daumé III (me@hal3.name)

Preliminary Results

No Seeds Seeds

10 20 30 40 50 60 No O/C Open/Close d

SLIDE 7

Symbol Pushing Slide 7 Hal Daumé III (me@hal3.name)

Preliminary Results: Open/Closed

No Rules Art<-N Prp->N Both 20 25 30 35 40 45 50 55 60 No Rules Art<-N Prp->N Both 20 25 30 35 40 45 50 55 60

NO SEEDS SEEDS

SLIDE 8

Symbol Pushing Slide 8 Hal Daumé III (me@hal3.name)

I'd like NLP to use more linguistics, but...

➢

Linguistic models are often developed without any reference to computation

➢

Many NLP students do not learn (or appreciate) much beyond other than Syntax I

➢