Previous Page  12 / 15 Next Page
Information
Show Menu
Previous Page 12 / 15 Next Page
Page Background

W H I T E P A P E R

© 2017 Persistent Systems Ltd. All rights reserved. 12

www.persistent.com

Advantages:

1. Ready-to-deploy containers with minimum configuration

a. Docker image

b. IBM containers for bluemix

2. Robust, well documented, developed at IBM

3. Ageneric solution offering several features

a. Transfer call to human agent, allow users to barge-in

b. Hang-up, music on hold, Control interaction via state variables embedded within Watson

conversation responses.

c. Latency auditing

d. Audio recording, DTMF support

Limitations:

1. Limited development and production use

2. Multiple tenancy not supported yet

3. Closed source, proprietary software, limited community and support

https://www.ibm.com/support/knowledgecenter/SS4U29/limitations.html

4. Enriching the response is not supported out of the box, however the feature is under development. A

workaround can be used where the WATSON_CONVERSATION_URL parameter of Voice gateway

could point to REST proxy which will forward the conversation request to Watson; then receive a

response andmodify it and pass back to Voice gateway.

https://developer.ibm.com/answers/questions/356108/customizing-voice-gateway-sip-orchestrator/

4. Challenges with Voice Agents

Although voice agents aid with customer service and alleviate a few pain points, there are certain challenges

that needs to be addressed to have a proper solution in place. As the systems get mature, many of the

challenges listed below will be overcome or fixed by right strategy. Note that some or all of the challenges below

also applies to the solutions listed above

1. No visible UI over telephone:

a. No visual cues to indicate when to start talking, when to stop taking and when voice-agent is thinking.

Example: When talking to Google Now or Siri, users can observe subtle visible cues indicating different

states of voice agents like when it is ready to listen, when it is listening, when it is processing the previous

utterance etc. AGUI also makes presenting some data formats, like tables, easier.

2.Accuracy of speech-to-text over traditional phone systems

a. Audio quality: traditional phone systems use G.711 codec and supports single channel, narrowband. The

sampling rate is lowwith of 8KHz sampling and 64 kbps bitrate.