Speech to text transcripts for ucce finesse

[deepaksharmainnovaccer]

Hi,

We have developed a third-party Call Control Handler that functions similarly to finesse desktop. This application works independently and uses finesse APIs to manage calls and respond to finesse events.

Currently, after a call ends, agents have to manually write a summary of the call and submit it through our application. This process is time-consuming and may result in agents overlooking important details because of the queued calls and limited wrap-up time.

To solve this problem, we are considering two options:

1. Implementing a speech-to-text solution that generates a real-time transcript of the call.
2. Creating a solution to record Jabber audio and converting it into text at the end of the call (though we're unsure about how to record Jabber calls).

We have researched solutions like Google Dialogflow and IBM Watson based on documentation and community posts. However, it appears that these solutions are better suited for integrating with finesse desktop or gadget, rather than our specific third-party application.

We are curious if there is a native solution available in UCCE or any other easy-to-integrate option that suits our use case.

Please provide your insights or suggestions. Thank you!

[david.macias]

Call summarization is pretty common now a days and will probably be the gold standard of what you would want to do. There are plenty of vendors who handle recording Verint, Calabrio and a few others. I would not use voice for your agents to enter wrap up data. If you do your homework right and keep a list of reason codes up to date that will be much easier on the agents. However, this is not something you just build and it works. QM/QA needs to be performed constantly to ensure the data you're getting is correct and worthwhile.

david

[deepaksharmainnovaccer]

Hi @david.macias ,

Thanks for the quick reply. In our use case, we can not give the agent more than 30 seconds for wrap-up because of the queued calls.

[Quigath]

I've done some work with transcription using AWS/Lex/Poly. The process is to send the recorded conversation to AWS, run the stt on it, then send the result back to the app. It was all custom so not sure how to integrate with the cisco side. Good luck!

[deepaksharmainnovaccer]

Hi @Quigath ,

Can you please tell me how are you recording the conversation?
Are you using any third party application?

I am thinking of using OpenAI's whisper package.

[Quigath]

I think we actually had an Avaya integration that got us the .wav files on disk. The process took a minute or two so it wasn't real-time.

[ravindra999]

Hi Deepak,

I have the similar requirement. Just wondering what solution did you use to get the transcript.  We have a client who uses B&S in their CRM and need the transcript attached to the case.  So we are looking at option of how to get transcript and show it to the agent. Appreciate any pointers you found during your research.

Thanks,
Ravi.

[david.macias]

This is a two part problem, first you need to record the conversation. Then get a hold of the wav file and have it transcribed. Do you currently have recordings and can access the wav file?

[mikehudson643]

Could you please explain the method you are using to record the conversation?
Do you use any applications from third parties?

I'm considering utilizing the whisper package from OpenAI.

[dhiarumu]

Hi

Are you looking for this?
https://www.cisco.com/c/en/us/td/docs/voice_ip_comm/cust_contact/contact_center/icm_enterprise/icm_enterprise_12_6_1/configuration/guide/ucce_b_features-guide-1261/rcct_m_agent-call-transcription_fg.html

Thanks