Summarization V2 baseline model for chat and voice

Agent Assist now supports a new V2 Summarization baseline model for voice and chat data. This model enables users to customize the content of the summary by selecting from the following predefined sections:

• Situation: What the customer needs help with or has question about.
• Action: What the agent does to help the customer.
• Resolution: Result of the customer service.
• Customer satisfaction: "Unsatisfied" if the customer is unsatisfied at the end of the conversation and "Satisfied" otherwise.
• Reason for cancellation: If the customer requests to cancel service. N/A otherwise.
• Entities: The key-value pairs of important entities extracted from the conversation.

Note: Summaries with customized sections are only available for the English language. For other supported languages in Summarization Baseline V2, summaries are produced as a single paragraph. Low-quality summaries might get deleted and returned empty.

Configure a conversation profile

A conversation profile configures a set of parameters that control the suggestions made to an agent during a conversation. The following steps create a ConversationProfile with a HumanAgentAssistantConfig object.

Create from console

We recommend creating a conversation profile using the Agent Assist console:

1. Input Display name for conversation profile.
2. Select Language from the list.
3. Select suggestion type Conversation summarization or Conversation summarization (voice).
4. Set Suggestion model type to Baseline model.
5. Set Baseline model version to 2.0.

6. If the selected language is English, select Output sections to be included in the summary.

   

Create from API

Note: If using Agent Assist's generateStatelessSummary API, this step can be skipped to instead provide the fields of the conversation profile within the request.

To create a conversation profile, do the following:

1. Call the create method on the ConversationProfile resource.
2. Provide a name for the new conversation profile.
3. Enter your Google Cloud project ID.
4. Enter the language code.
5. Enter your model ID. To use the V2 baseline model, don't specify the model ID. Instead, set baseline_model_version to 2.0.
6. If entered language is English, in the CreateConversationProfileRequest for the query configuration, specify sections to be included in the summary. The default sections are SITUATION, ACTION, and RESOLUTION if no section is specified.
7. In the CreateConversationProfileRequest for the suggestion feature, specify CONVERSATION_SUMMARIZATION for chat or CONVERSATION_SUMMARIZATION_VOICE for voice.

Baseline model example for voice

The following is a JSON example that uses a baseline summarization model for voice:

{
  "displayName": "CONVERSATION_PROFILE_NAME",
  "humanAgentAssistantConfig": {
    "humanAgentSuggestionConfig": {
      "featureConfigs": [{
        "suggestionFeature": {
          "type": "CONVERSATION_SUMMARIZATION_VOICE"
        },
        "queryConfig": {
          "sections": {
            "sectionTypes": "SITUATION",
            "sectionTypes": "ACTION",
            "sectionTypes": "RESOLUTION",
            "sectionTypes": "REASON_FOR_CANCELLATION",
            "sectionTypes": "CUSTOMER_SATISFACTION",
            "sectionTypes": "ENTITIES"
          }
        },
        "conversationModelConfig": {
          "baselineModelVersion": "2.0",
        }
      }]
    }
  },
  "languageCode": "en-US"
}

The response is a ConversationProfile object containing the conversation profile name:

{
  "name": "projects/PROJECT_ID/locations/global/conversationProfiles/CONVERSATION_PROFILE_ID",
  "displayName": "CONVERSATION_PROFILE_NAME",
  "automatedAgentConfig": {
  },
  "humanAgentAssistantConfig": {
    "notificationConfig": {
    },
    "humanAgentSuggestionConfig": {
      "featureConfigs": [{
        "suggestionFeature": {
          "type": "CONVERSATION_SUMMARIZATION_VOICE"
        },
        "queryConfig": {
          "sections": {
            "sectionTypes": [
              "SITUATION",
              "ACTION",
              "RESOLUTION",
              "REASON_FOR_CANCELLATION",
              "CUSTOMER_SATISFACTION",
              "ENTITIES"
            ]
          }
        },
        "conversationModelConfig": {
          "baselineModelVersion": "2.0",
        }
      }]
    },
    "messageAnalysisConfig": {
    }
  },
  "languageCode": "en-US",
  "createTime": "2023-07-06T21:06:46.841816Z",
  "updateTime": "2023-07-06T21:06:46.841816Z",
  "projectNumber": "344549229138"
}

Baseline model example for chat

The following is a JSON example that uses a baseline summarization model for chat:

{
  "displayName": "CONVERSATION_PROFILE_NAME",
  "humanAgentAssistantConfig": {
    "humanAgentSuggestionConfig": {
      "featureConfigs": [{
        "suggestionFeature": {
          "type": "CONVERSATION_SUMMARIZATION"
        },
        "queryConfig": {
          "sections": {
            "sectionTypes": "SITUATION",
            "sectionTypes": "ACTION",
            "sectionTypes": "RESOLUTION",
            "sectionTypes": "REASON_FOR_CANCELLATION",
            "sectionTypes": "CUSTOMER_SATISFACTION",
            "sectionTypes": "ENTITIES"
          }
        },
        "conversationModelConfig": {
          "baselineModelVersion": "2.0",
        }
      }]
    }
  },
  "languageCode": "en-US"
}

The response is a ConversationProfile object containing the conversation profile name:

{
  "name": "projects/PROJECT_ID/locations/global/conversationProfiles/CONVERSATION_PROFILE_ID",
  "displayName": "CONVERSATION_PROFILE_NAME",
  "automatedAgentConfig": {
  },
  "humanAgentAssistantConfig": {
    "notificationConfig": {
    },
    "humanAgentSuggestionConfig": {
      "featureConfigs": [{
        "suggestionFeature": {
          "type": "CONVERSATION_SUMMARIZATION"
        },
        "queryConfig": {
          "sections": {
            "sectionTypes": [
              "SITUATION",
              "ACTION",
              "RESOLUTION",
              "REASON_FOR_CANCELLATION",
              "CUSTOMER_SATISFACTION",
              "ENTITIES"
            ]
          }
        },
        "conversationModelConfig": {
          "baselineModelVersion": "2.0",
        }
      }]
    },
    "messageAnalysisConfig": {
    }
  },
  "languageCode": "en-US",
  "createTime": "2023-07-06T21:06:46.841816Z",
  "updateTime": "2023-07-06T21:06:46.841816Z",
  "projectNumber": "344549229138"
}

Handle conversations at runtime

Conversations are handled the same at runtime whether you're using the baseline model or a custom model. See the Summarization custom model documentation for details.