Wednesday, July 19, 2006
#2 (from RTE 1) Summary
http://www.cs.biu.ac.il/~glikmao/rte05/akhmatova.pdf
Kullanilan kaynaklar:
Text (T) ve Hypothesis (H) 'te proposition'lar detect ediliyor. Eger H'teki her bir proposition icin T'de bir matching proposition var ise, T entails H.
- Asagidaki sekilde logical formlar kullaniyorlar:
3 tip object var: Subj(x), Obj(x), Pred(x)
and a meaning attaching element: iq(x,)
ORNEK:
"Somali capital" -- Subj(x) & iq(x, 'capital') & attr(x, y) & Subj(y) & iq(y, 'somali')
"a zoo in Berlin" -- Obj(x) & iq(x, 'zoo') & prep(x, y) & Obj(y) & iq(y, 'Berlin')
- Wordnet 'teki relation'lardan logic formula'lar generate ediyorlar.
ORNEK:
Semantic synonymy is expressed as an equivalence:
Kullanilan kaynaklar:
- Wordnet
- Link Parser
- OTTER
- first-order logic forms
- Wordnet Relatedness Algorithm
Text (T) ve Hypothesis (H) 'te proposition'lar detect ediliyor. Eger H'teki her bir proposition icin T'de bir matching proposition var ise, T entails H.
- Asagidaki sekilde logical formlar kullaniyorlar:
3 tip object var: Subj(x), Obj(x), Pred(x)
and a meaning attaching element: iq(x,
ORNEK:
"a zoo in Berlin" -- Obj(x) & iq(x, 'zoo') & prep(x, y) & Obj(y) & iq(y, 'Berlin')
- Wordnet 'teki relation'lardan logic formula'lar generate ediyorlar.
ORNEK:
Semantic synonymy is expressed as an equivalence:
iq(x, 'serve') <-> iq(x, 'dish')
and hypernymy as:
iq(x, 'serve') -> iq(x, 'provide')
- Bazi lexical relation rule'lari store ediyorlar:
ORNEK:
- Bazi lexical relation rule'lari store ediyorlar:
ORNEK:
Ax (iq(x, 'is') <-> iq(x, 'be'))
- They describe syntactical equivalence by means of additional logic rules:
is
A y z z1 y77 z77 ((Pred(z1) & iq(z1, 'be') & Obj(y) & iq(y, y77) & prep(y, z) & Obj(z) & iq(z, z77))))
WordNet Relatedness:
- Iki kelime arasindaki path ne kadar uzun ise relatedness o kadar azaliyor.
- Path icerisindeki kelimelerin sense'leri ne kadar az ise, o kadar relatedness artiyor. (two words connected with a verb print are more close to each other, than the words connected via make; print has 4 senses, make has 49)
- Iki sense arasinda ne kadar cok farkli path varsa o kadar related.
ORNEK:
provide -- give (verb chain)
1. provide#1(7)[2259805] -- hyperonym -- give#3(44)[2136207]
...
7. provide#6(7)[2155855] -- hyperonym -- support#2(11)[2155507] -- hyperonym -- give#3(44)[2136207]
maximum path length -- 3
relatedness score: 0.3623 => Ax (iq(x, 'provide') -> iq(x, 'give'))
Bir T-H pair icin ORNEK:
T: Coffee boosts energy and proveides health benefits.
H: Coffee gives health benefits.
Propositions for T:
- Coffee boosts energy.
- Coffee provides health benefits.
Proposition for H:
- Coffee gives health benefits.
H'teki butun proposition'lar T'deki proposition'lardan bir tanesi ile match etmesi gerekiyor.
for (Coffee gives health benefits)
[FALSE]Coffee boosts energy => Coffee gives health benefits (FALSE)
[TRUE]Coffee provides health benefits => Coffee gives health benefits, as provides => gives, coffee = coffee, health benefits = health benefits
Performance of the System:
accuracy / cws = 0.5067 / 0.5188
"Be X of Y -> X Y" kullanarak verilen dogru cevaplara ORNEK:
T: The man is a director of the company
H: The man rules the company.
Yanlis cevablara ORNEK:
T: The man came to the park by car.
H: The man came to a car park.
- They describe syntactical equivalence by means of additional logic rules:
ORNEK:
"Be X of Y -> X Y" (director of the firm -> direct a firm)is
A y z z1 y77 z77 ((Pred(z1) & iq(z1, 'be') & Obj(y) & iq(y, y77) & prep(y, z) & Obj(z) & iq(z, z77))))
WordNet Relatedness:
- Iki kelime arasindaki path ne kadar uzun ise relatedness o kadar azaliyor.
- Path icerisindeki kelimelerin sense'leri ne kadar az ise, o kadar relatedness artiyor. (two words connected with a verb print are more close to each other, than the words connected via make; print has 4 senses, make has 49)
- Iki sense arasinda ne kadar cok farkli path varsa o kadar related.
ORNEK:
provide -- give (verb chain)
1. provide#1(7)[2259805] -- hyperonym -- give#3(44)[2136207]
...
7. provide#6(7)[2155855] -- hyperonym -- support#2(11)[2155507] -- hyperonym -- give#3(44)[2136207]
maximum path length -- 3
relatedness score: 0.3623 => Ax (iq(x, 'provide') -> iq(x, 'give'))
Bir T-H pair icin ORNEK:
T: Coffee boosts energy and proveides health benefits.
H: Coffee gives health benefits.
Propositions for T:
- Coffee boosts energy.
- Coffee provides health benefits.
Proposition for H:
- Coffee gives health benefits.
H'teki butun proposition'lar T'deki proposition'lardan bir tanesi ile match etmesi gerekiyor.
for (Coffee gives health benefits)
[FALSE]Coffee boosts energy => Coffee gives health benefits (FALSE)
[TRUE]Coffee provides health benefits => Coffee gives health benefits, as provides => gives, coffee = coffee, health benefits = health benefits
Performance of the System:
accuracy / cws = 0.5067 / 0.5188
"Be X of Y -> X Y" kullanarak verilen dogru cevaplara ORNEK:
T: The man is a director of the company
H: The man rules the company.
Yanlis cevablara ORNEK:
T: The man came to the park by car.
H: The man came to a car park.
# posted by Bengi Mizrahi @ 11:06 AM 
