演習問題 2014.7.22 系列モデリング Sequential data modeling
名前
Name
No.
n1 n2 n3 n4 n5 n6 n7
以下のように品詞の異なる系列を設定します。
We set the different tag sequences as follows.
タグ系列
(tag sequence)
Time
flies
like
an
arrow
1
N
V
P
D
N
2
A
N
V
D
N
3
V
N
P
D
N
※ 系列の意味はそれぞれ:N=名詞 ・V=動詞 ・P=前置詞 ・D=冠詞 ・A=形容詞
The meaning of each tag is: N=Noun, V=Verb, P=Preposition, D=Definite A= Adjective
<意味>
タグ列1: 光陰矢のごとし
タグ列2: 矢が好きな時蝿
タグ列3: 矢のような蝿の時間を計れ
<mean>
tag1: Time flies like an arrow
tag2: The time flies which likes an arrow
tag3: Measure time of flies such as a arrow
Submit to : B7F, Augmented Human Communication Lab, Submit Box
Deadline : 7/29 AM9:00 , Questions to TA (Mailto: [email protected])
1
演習問題 2014.7.22 系列モデリング Sequential data modeling
Q1. タグ列1の遷移  T ・ 生成  E ・ 大文字  CAPS それぞれの素性関数を求めよ。全12個あり。
Find the transition, emission and caps features of the tag sequence 1. There are 12 features in total.
Answer
2
演習問題 2014.7.22 系列モデリング Sequential data modeling
Q2. それぞれの重みを以下のように設定する。
We set the weight vectors as follows.
wT,<S>,N = 2
wT,D,N = n2
wT,V,D = 2
wE,N,"Time" = 2
wE,N,"arrow" =1
wE,V,"Time" = n7
wCAPS,N = n 5
wT,N,V = n1
wT,N,</S> = n6
wT,<S>,V = 2
wE,V," flies" =1
wT ,V,P =1
wT ,<S>,A = 3
wE,A,"Time" =1
wT,V,N = n7
wE,P,"like" =1
wE,N," flies" = n4
wCAPS,A =1
wCAPS ,V  1
wT ,P,D =1
wT,A,N = 4
wT,N,P = n3
wE,D,"an" =1
wE,V,"like" =1
タグ列1のスコアを求めよ。
Find score of tag sequence 1.
Score of tag sequence 1 :
3
演習問題 2014.7.22 系列モデリング Sequential data modeling
Q3. タグ列2とタグ列3のスコアを求めよ。 また,どちらが最大になるか。
Find the score of tag sequence 2 and tag sequence 3. And which tag sequence is maximum?
Score of tag sequence 2 :
Score of tag sequence 3 :
Maximum tag sequence:
4
演習問題 2014.7.22 系列モデリング Sequential data modeling
Q4. 正解の素性ベクトルとスコア最大の素性ベクトルを利用し、重みの更新を行え。
Update the weights vector using correct features and features which is maximum score.
ただし、タグ1列を正解の素性ベクトルとする。
We assume that tag sequence 1 is correct features.
Answer
wT,<S>,N
wT ,N,V
wE,N,"Time"
wE,V," flies"
wT ,V,P
wT ,P,D
wT ,D,N
wT ,N,</S>
wT ,<S>,A
wT ,A,N
wT ,V , D
wE,P,"like"
wE,D,"an"
wCAPS,N
wCAPS,A
wCAPS,V
wE,N,"arrow"
wE,A,"Time"
wE,N," flies"
wE,V,"like"
wE,V,"Time"
wT ,<S>,V
wT ,V,N
wT ,N,P
5
演習問題 2014.7.22 系列モデリング Sequential data modeling
<おまけ問題(Extra question)>
品詞推定以外の問題を、構造化パーセプトロンを使って解くとしたらどのような入出力で
どのような素性を利用したらよいか考察せよ。
What kind of input, output and features should we use to solve the problem using the structured perceptron.
Answer your idea.
6