Information Technology Reference
In-Depth Information
We found that the chat in Second Life was similar to the chats on the Internet.
The users did not care much about grammar and spelling. When the subjects
used the portable translation system, the other party did not answer in consid-
eration of such a special situation.
5 Experiment 2: Fixed Multilingual Chat Communication
System
5.1 Experimental Procedure
We carried out the experiment using the fixed multilingual chat communication
system. The purpose of the experiment was to investigate what problems oc-
curred during the use of the fixed multilingual chat communication system in
Second Life.
In the conversation experiment, we verified the following items:
-
Can you carry out a serious conversation using this system?
-
Does the meaning of the other party's sentence change because of the ma-
chine translation?
-
How can we verify that the machine translation function is effective?
From the result of the experiment using the portable system, it was concluded
that most of the conversation was not established. We thought that this was
because of the following two reasons:
-
The other party did not understand the purpose of the conversation.
-
The conversation was not meant for machine translation.
Therefore, we gave the subject a purpose “Question your partner by using ma-
chine translation.” The subjects were seven Second Life users whose native lan-
guage was English. We recruited the experimental subjects from the ocial Web
page of Second Life. We told the participating subjects that they would be paid
10L$'30L$ in Second Life currency as a reward.
The procedure of the experiment was as follows:
1. We guided the subject to the translation chat room.
2. We told the subject that this was an experiment to verify the effectiveness
of machine translation.
3. We asked the subject to ask a Japanese interlocutor 10 different questions.
4. The Japanese interlocutor requested rephrasing when the question could not
be understood from the translation result.
5. When the translation of the rephrased sentence also did not make much
sense, the Japanese interlocutor asked the subject to change the question.
The Japanese interlocutor was one of the authors. In the experiment, the subjects
did not use back translation. They could only see their input sentences and the
translated sentences. We screen-captured the entire experiment. We recorded all
the sentences that were exchanged during the chat.
