Information Technology Reference
In-Depth Information
Table 3.
Judgment result of the Japanese interlocutor's understanding of the subjects'
question
Question of subject
Apprehensible Inapprehensible Average number
of characters
Subjects
Average
number
of words
A
7
3
74.8
14.6
B
10
0
39.5
7.6
C
10
0
31.9
6.2
D
8
2
53.1
10.0
E
9
1
57.9
10.7
F
9
1
46.7
9.9
G
10
0
66.2
13.1
Average
9.0
1.0
52.9
10.3
5.2 Result and Discussion
Table 3 shows the Japanese interlocutor's understanding of the subjects' ques-
tions. The Japanese interlocutor judged whether a translated question was un-
derstandable. The subjects were asked to rephrase questions that the Japanese
interlocutor did not understand. From Table 3, it could be concluded that the
translation result for many of the questions was understandable.
Some of the problems that hindered accurate translation, and hence the overall
chat communication, were as follows:
1. The subjects hardly used clipped forms in the chat. We had explained to
the subjects that the experiment was about verifying the effectiveness of
machine translation in advance. However, we did not explain the features of
machine translation and the problems caused by the use of clipped forms.
2. The subjects often made spelling mistakes. Therefore, the machine transla-
tion failed. A couple of examples are shown below.
Input sentence:
Who are your favorate singers/music groups?
Trans lation resul t:
favorate/yO[vNH
(The spelling of “favorate ”is incorrect.)
Input sentence:
It is a very well desgined welcome area.
Trans lation resul t:
desginedGAB
(The spelling of “desgined ”is incorrect.)
3. When the Japanese interlocutor said that he did not understand the trans-
lated sentence, the subjects rephrased the sentence. The subject followed the
following rephrasing method:
-
The subject numbered each sentence.
-
The subject supplemented the meaning by using parentheses.
Some of the examples are as follows:
First input sentence:
I am making up questions, it is ok. Next question: This
skirt comes in Copy/No Mod/No Trans, can you make the skirt in Copy/Mod/No
Trans? I am small, so I have to edit the skirt.
Second input sentence:
(1) I am making up questions, it is ok. (2) This skirt
