Cloud Workflow

Call Center with Twilio, without Telephone Machine (‘Please wait for a few seconds.’ in Japanese)

Auto Answer in Japanese, corresponding to Twilio's official launch of service in Japan.

The first article of this series, was about the implementation of automatic Starting a Process by incoming phone call. But because it was receiving calls on the phone, I made it to receive on the browser in the second article, so to look cool.
(The articles of so far are HERE.)

Now this time, I am going to make the Auto-answering in Japanese!

… You might say ‘What in the world is that?’, but till today the Text-to-speech synthesis of the Twilio does not support Japanese language, so my system always have been answering in English. But now Twilio started supporting Japanese with their Text-to-speech synthesis, so I am going to brush up the system to answer in Japanese.

This time, the point of improvement is simple. I have modified the content of the XML (TwiML) which would be returned to the HTTP Request to the pre-registered URL when a phone call comes in. The modified point is the section of <Say> in line 2.

<Response>
  <Say language="ja-jp" voice="woman">しばらくお待ちください。</Say>
  <Dial record="true" action="http://questetra-twilio.appspot.com/callend" method="GET">
    <Client>support</Client>
  </Dial>
</Response>

It will make text-to-speech synthesis in Japanese by specifying ‘ja-jp’ which represents the Japanese language in the attribute of language. The attribute of ‘voice’ is for setting the answering in a male or female voice. It is set as male as default, and modify if you would like to change.

Explaining other parts, <Dial> tag is a command to call forwarding. The Attribute of ‘record=”true”‘ for recording. In the ‘action’ attribute, specifying a URL for HTTP request at the call end. <Client> tag is for specifying destination of the call forwarding.

ell going back to the auto-answer section, I am concerned about how good the ‘text-to-speech synthesis in Japanese’ is. I dared to try to copy & paste an article in a newspaper of the economy. I expected it would be messed up, but it had read almost correctly. The text-to-speech synthesis technology seems to have evolved considerably while I did not pay attention.

However, I had the impression that it ‘has some accents’ for some reason. I never felt ‘have accents’ before, though I often heard the voice of ‘machine-like’ in IVR systems. You may think it is not good to ‘have accents’. In that case, it is also possible to play back the voice recorded in advance.

<Response>
  <Play>http://questetra-twilio.appspot.com/hello.mp3</Play>
  <Dial record="true" action="http://questetra-twilio.appspot.com/callend" method="GET">
    <Client>support</Client>
  </Dial>
</Response>

In such a case, use <Play> tag instead of <Say>. The sound file at the specified URL in the <Play> will be played.

 

That is all for this time. Twilio has started providing an official service, which will open a way to Full-scale use in Japan. Next time, I would like to work on how to convey Touch-tone entry data to a Business Process.

Recommendations
Prev article - 15. Ambitions Three changes caused by the introduction of Enterprise Social Networking
Next article - 15. Ambitions What is This Thing Called 'Business Process'?
Another article - Hatanaka Akihiro Call Center with Twilio, without Telephone Machine (Receiving Phone calls on the Browser)

Archive

 RSS