1. Trang chủ
  2. » Kinh Doanh - Tiếp Thị

A Markov Logic Approach to Bio-Molecular Event Extraction potx

9 335 0

Đang tải... (xem toàn văn)

THÔNG TIN TÀI LIỆU

Thông tin cơ bản

Định dạng
Số trang 9
Dung lượng 389,98 KB

Nội dung

Proceedings of the Workshop on BioNLP: Shared Task, pages 41–49, Boulder, Colorado, June 2009. c 2009 Association for Computational Linguistics ∗† ∗† ∗¶ †‡§ ∗ † ¶ ‡ § {sebastian,chun,takagi}@dbcls.rois.ac.jp tsujii@is.s.u-tokyo.ac.jp 41 eventType (e, t) ∧ role (e, a, r) ∧ event (a) ⇒ regT ype (t)        1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 42 e (i, t, A) i e t e A (a, r) a r a e P L E P = {4, 7} , L = ∅ E = {e 13 , e 14 , e 15 } e 15 = (5, , {(4, )}) e 14 = (2, , {(e 15 , ) , (7, )}) e 13 = (1, , {(e 14 , )}) E E (L, C) L (i, j, r) C (i, t) P L E E (L, C) C = {(1, ) , (2, ) , (5, )} / * returns all clues C and links L given by the events in E * / 1 function eventsT oLinks (E): 2 C ← ∅, L ← ∅ 3 for each event (i, t, A) ∈ E do 4 C ← C∪{(i, t)} 5 for each argument (a, r) ∈ A do 6 if a is an event (i  , t  , A  ) do 7 L ← L∪{(i, i  , r)} with a = (i  , t  , A  ) 8 else 9 L ← L ∪ {(i, a, r)} 10 return (C, L) L = {(1, 2, ) , (2, 5, ) , (2, 7, ) , (5, 4, )} . L c c L c c C L E 43 / * returns all events E specified by clues C and links L * / 1 function linksT oEvents (C, L) 2 return S (i,t)∈C resolve (i, C, L) / * returns all events for the given token i * / 1 function resolve (i, C, L) 2 if no t with (i, t) ∈ C return {i} 3 t ← type (i, C) 4 if t = binding return {(i, t, A)} with 5 A = {(a, r) | (i, a, r) ∈ L} 6 R i ← {r  |∃a : (i, a, r) ∈ L} 7 for each role r ∈ R i do 8 A r ← {a| (i, a, r) ∈ L} 9 B r ← S a∈A r {(resolve (a) , r)} 10 return S A∈expand ( B r 1 , ,B r n ) {(i, t, A)} / * returns all possible argument sets for B r 1 , . . . , B r n * / 1 function expand (B r 1 , . . . , B r n ) 2 if n = 1 return B r n 3 return S a∈B r 1 S A∈expand ( B r 2 , ,B r n ) {(a, r 1 )} ∪ A {word (1, ) , eventT ype (1, ) , role(1, 2, ), event(2), . . .} {word (1, ) , eventT ype (1, ) , role(1, 2, ), event(2), . . .} M (φ, w) φ w M p (y) = 1 Z exp    (φ,w)∈M w  c∈C φ f φ c (y)   y C φ φ f φ c y φ c Z 44 F P + αF N α = 0.01 i i t i i j r L C L p i j parser ↓ ↓ ↓↓ i w i s i p i w i s i d i p i j d parser parser i j p p i j path dep (1, 2, , ) ∧ word (1, ) ⇒ eventType (2, ) dep(h, i, d, parser) ∧ word (h, +w) ⇒ eventType(i, +t) (w, t) event (i) 45 T (i, +t) ⇒ event (i) T T word (i, +t) ⇒ event (i) genia (i, p) ⇒ event (i) p role (i, j, +r) dict (i, +d i ) ∧ dict (j, +d j ) ⇒ role (i, j, +r) path (i, j, +p, +parser) ⇒ role (i, j, +r) i j j i P (i, j, +p, +parser) ∧ T (i, +t) ⇒ role (i, j, +r) P T T (i, +t) T (i) 46 event (i) ⇒ ∃t.eventT ype (i, t) eventT y pe (i, t) ⇒ event (i) eventT y pe (i, t) ∧ t = o ⇒ ¬eventT ype (i, o) ¬site (i) ∨ ¬event (i) role (i, j, r) ⇒ event (i) j r i i role (i, j, r 1 ) ∧ r 1 = r 2 ⇒ ¬role (i, j, r 2 ) eventT y pe (e, t) ∧ role (e, a, r) ∧ event (a) ⇒ regT ype (t) role (i, j, r) ∧ taskOne (r) ⇒ event (j) ∨ protein (j) role (i, j, r) ∧ taskTwo (r) ⇒ site (j) site (j) ⇒ ∃i, r.role (i, j, r) ∧ taskTwo (r) event (i) ⇒ ∃j.role (i, j, ) eventT y pe (i, t) ∧ ¬allowed (t, r) ⇒ ¬role (i, j, r) role (i, j, r 1 ) ∧ k = i ⇒ ¬r ole (k, j, r 2 ) j < k ∧ i < j ∧ role (i, j, r 1 ) ⇒ ¬role (i, k, r 2 ) 47 48 49 . Shared Task, pages 41–49, Boulder, Colorado, June 2009. c 2009 Association for Computational Linguistics ∗† ∗† ∗¶ †‡§ ∗ † ¶ ‡ § {sebastian,chun,takagi}@dbcls.rois.ac.jp tsujii@is.s.u-tokyo.ac.jp 41 eventType. for each argument (a, r) ∈ A do 6 if a is an event (i  , t  , A  ) do 7 L ← L∪{(i, i  , r)} with a = (i  , t  , A  ) 8 else 9 L ← L ∪ {(i, a, r)} 10

Ngày đăng: 23/03/2014, 13:20

TỪ KHÓA LIÊN QUAN

TÀI LIỆU CÙNG NGƯỜI DÙNG

TÀI LIỆU LIÊN QUAN

w