Genesys Cloud - Main

I'm building a call flow where the caller provides a spoken input containing a full USA address. My initial idea was to use the Collect Input action to capture the entire speech input, convert it to text, and store it in a variable. Then, I planned to use external logic to parse the text into street number, street name, city, state, and ZIP code.

However, I've hit a roadblock: it seems Genesys Cloud doesn't provide a way to collect free-form speech input and store it as text in a variable.

Has anyone implemented something similar? Any suggestions or best practices for achieving this requirement?

Hi Mohamed,

I think the Collect Input action is more for collecting DTMF or Specific Terms like "Account Ballance" as opposed to large text.

You could use a bot flow with slots for Street Number, Street Name, City etc to capture the information.  

Maybe others in the community have a better way of doing this though.

I'm building a call flow where the caller provides a spoken input containing a full USA address. My initial idea was to use the Collect Input action to capture the entire speech input, convert it to text, and store it in a variable. Then, I planned to use external logic to parse the text into street number, street name, city, state, and ZIP code.

However, I've hit a roadblock: it seems Genesys Cloud doesn't provide a way to collect free-form speech input and store it as text in a variable.

Has anyone implemented something similar? Any suggestions or best practices for achieving this requirement?

Challenge with Speech Input for Multiple Slot Types in a Single Utterance

We've been working on improving the approach to capture speech input that includes business name, U.S. address, and purpose of the call within a single utterance. However, the Virtual Agent is struggling to correctly identify all Slot Types-particularly the Address SlotType, which fails consistently during live speech input.

Even though NLU and regex validation tests pass in isolation, the issue occurs when customers provide free-form speech. For example:

In this case:

• BusinessName and CallerIntent SlotTypes are assigned correctly.
• AddressSlotType is not being recognized.

Since we allow callers to speak naturally without restrictions, they often provide all slot values in one sentence. This flexibility seems to be causing recognition challenges for the address component.

------------------------------
Mohamed Salim Nagoor Ghani

Original Message:
Sent: 11-13-2025 11:08
From: Samuel Jillard
Subject: Capture Caller Speech Input - USA Address

Hi Mohamed,

I think the Collect Input action is more for collecting DTMF or Specific Terms like "Account Ballance" as opposed to large text.

You could use a bot flow with slots for Street Number, Street Name, City etc to capture the information.  

Maybe others in the community have a better way of doing this though.

------------------------------
Sam Jillard
Online Community Manager/Moderator
Genesys - Employees

Original Message:
Sent: 11-12-2025 23:53
From: Mohamed Salim Nagoor Ghani
Subject: Capture Caller Speech Input - USA Address

I'm building a call flow where the caller provides a spoken input containing a full USA address. My initial idea was to use the Collect Input action to capture the entire speech input, convert it to text, and store it in a variable. Then, I planned to use external logic to parse the text into street number, street name, city, state, and ZIP code.

However, I've hit a roadblock: it seems Genesys Cloud doesn't provide a way to collect free-form speech input and store it as text in a variable.

Has anyone implemented something similar? Any suggestions or best practices for achieving this requirement?

#ArchitectandDesign
#ConversationalAI(Bots,VirtualAgent,etc.)
#Implementation
#Telephony

------------------------------
Mohamed Salim
------------------------------

Quick questions @Mohamed Salim Nagoor Ghani:

Have you tried using an AI-Powered Slot Type for the AddressSlotType within your Virtual Agent? If so, can you tell where it's failing when you replay your phone call using flow execution history? Also worth keeping the Virtual Agent slot authoring recommendations and limitations in mind.

------------------------------
Brian T. Jones | Ascension | Senior Specialist - Technology

Original Message:
Sent: 11-21-2025 05:41
From: Mohamed Salim Nagoor Ghani
Subject: Capture Caller Speech Input - USA Address

Challenge with Speech Input for Multiple Slot Types in a Single Utterance

We've been working on improving the approach to capture speech input that includes business name, U.S. address, and purpose of the call within a single utterance. However, the Virtual Agent is struggling to correctly identify all Slot Types-particularly the Address SlotType, which fails consistently during live speech input.

Even though NLU and regex validation tests pass in isolation, the issue occurs when customers provide free-form speech. For example:

In this case:

• BusinessName and CallerIntent SlotTypes are assigned correctly.
• AddressSlotType is not being recognized.

Since we allow callers to speak naturally without restrictions, they often provide all slot values in one sentence. This flexibility seems to be causing recognition challenges for the address component.

------------------------------
Mohamed Salim Nagoor Ghani

Original Message:
Sent: 11-13-2025 11:08
From: Samuel Jillard
Subject: Capture Caller Speech Input - USA Address

Hi Mohamed,

I think the Collect Input action is more for collecting DTMF or Specific Terms like "Account Ballance" as opposed to large text.

You could use a bot flow with slots for Street Number, Street Name, City etc to capture the information.  

Maybe others in the community have a better way of doing this though.

------------------------------
Sam Jillard
Online Community Manager/Moderator
Genesys - Employees

Original Message:
Sent: 11-12-2025 23:53
From: Mohamed Salim Nagoor Ghani
Subject: Capture Caller Speech Input - USA Address

I'm building a call flow where the caller provides a spoken input containing a full USA address. My initial idea was to use the Collect Input action to capture the entire speech input, convert it to text, and store it in a variable. Then, I planned to use external logic to parse the text into street number, street name, city, state, and ZIP code.

However, I've hit a roadblock: it seems Genesys Cloud doesn't provide a way to collect free-form speech input and store it as text in a variable.

Has anyone implemented something similar? Any suggestions or best practices for achieving this requirement?

#ArchitectandDesign
#ConversationalAI(Bots,VirtualAgent,etc.)
#Implementation
#Telephony

------------------------------
Mohamed Salim
------------------------------

Challenge with Speech Input for Multiple Slot Types in a Single Utterance

We've been working on improving the approach to capture speech input that includes business name, U.S. address, and purpose of the call within a single utterance. However, the Virtual Agent is struggling to correctly identify all Slot Types-particularly the Address SlotType, which fails consistently during live speech input.

Even though NLU and regex validation tests pass in isolation, the issue occurs when customers provide free-form speech. For example:

In this case:

• BusinessName and CallerIntent SlotTypes are assigned correctly.
• AddressSlotType is not being recognized.

Since we allow callers to speak naturally without restrictions, they often provide all slot values in one sentence. This flexibility seems to be causing recognition challenges for the address component.

Hi Mohamed,

I think the Collect Input action is more for collecting DTMF or Specific Terms like "Account Ballance" as opposed to large text.

You could use a bot flow with slots for Street Number, Street Name, City etc to capture the information.  

Maybe others in the community have a better way of doing this though.

I'm building a call flow where the caller provides a spoken input containing a full USA address. My initial idea was to use the Collect Input action to capture the entire speech input, convert it to text, and store it in a variable. Then, I planned to use external logic to parse the text into street number, street name, city, state, and ZIP code.

However, I've hit a roadblock: it seems Genesys Cloud doesn't provide a way to collect free-form speech input and store it as text in a variable.

Has anyone implemented something similar? Any suggestions or best practices for achieving this requirement?

This would be a good case for Virtual Agents with Guides. Allowing the guide to parse the utterance into multiple variables.

Challenge with Speech Input for Multiple Slot Types in a Single Utterance

We've been working on improving the approach to capture speech input that includes business name, U.S. address, and purpose of the call within a single utterance. However, the Virtual Agent is struggling to correctly identify all Slot Types-particularly the Address SlotType, which fails consistently during live speech input.

Even though NLU and regex validation tests pass in isolation, the issue occurs when customers provide free-form speech. For example:

In this case:

• BusinessName and CallerIntent SlotTypes are assigned correctly.
• AddressSlotType is not being recognized.

Since we allow callers to speak naturally without restrictions, they often provide all slot values in one sentence. This flexibility seems to be causing recognition challenges for the address component.

Hi Mohamed,

I think the Collect Input action is more for collecting DTMF or Specific Terms like "Account Ballance" as opposed to large text.

You could use a bot flow with slots for Street Number, Street Name, City etc to capture the information.  

Maybe others in the community have a better way of doing this though.

I'm building a call flow where the caller provides a spoken input containing a full USA address. My initial idea was to use the Collect Input action to capture the entire speech input, convert it to text, and store it in a variable. Then, I planned to use external logic to parse the text into street number, street name, city, state, and ZIP code.

However, I've hit a roadblock: it seems Genesys Cloud doesn't provide a way to collect free-form speech input and store it as text in a variable.

Has anyone implemented something similar? Any suggestions or best practices for achieving this requirement?