Filtering

You can filter conversations based on information gleaned directly from their characteristics or from attributes of operations that those conversations have been through.

Conversation filters

Conversations can be filtered on one or more of the characteristics in the following table. These filters are useful for narrowing down a search among large conversation data sets. The "Filtering" section of the List Conversations page lists the permissions you need to apply the following filters.

Note: When using the double quotes (") in a curl command or a URL, you should escape them with a backslash and no extra space after (").

Filter term	Description	Reference code
Agent ID	String for a unique agent	agent_id="some agent_id" quality_metadata.agent_info.agent_id:"some agent_id"
Agent team	String for an agent team	agent_team="some agent_team" quality_metadata.agent_info.teams:"some agent_team"
Agent type	Human or automated agent	quality_metadata.agent_info.agent_type:"HUMAN_AGENT" quality_metadata.agent_info.agent_type="AUTOMATED_AGENT"
Analysis create time	Time when the analysis was created	latest_analysis.create_time>"2023-12-31T16:00:10-08:00"
Analysis status	Whether the conversation has been analyzed	Has been analyzed: latest_analysis:"*" Has not been analyzed: -latest_analysis:"*"
Conversation channel	Medium of the conversation: PHONE_CALL or CHAT	medium="PHONE_CALL"
Conversation create time	Time when the conversation was imported into Insights	create_time>"1969-12-31T16:00:10-08:00" create_time<="1969-12-31T16:00:20-08:00"
Conversation label	Label or tag assigned to a conversation	labels:"key_only" labels.key_1:"value_1"
Conversation start time	Time at which the conversation began	start_time>"1969-12-31T16:00:10-08:00" start_time>"1969-12-31T16:00:20-08:00"
Conversation update time	Time when the conversation was last updated	update_time>"1969-12-31T16:00:10-08:00" update_time<="1969-12-31T16:00:20-08:00"
CSAT score	Customer satisfaction score	quality_metadata.customer_satisfaction_rating>"1" quality_metadata.customer_satisfaction_rating<"5"
Custom highlight (also called Phrase matchers)	Custom highlights to help flag certain phrases	latest_analysis.analysis_result.call_analysis_metadata.phrase_matchers.id:"add_one_positive_custom_highlight" latest_analysis.analysis_result.call_analysis_metadata.phrase_matchers.id:"multiple positive custom highlights" -latest_analysis.analysis_result.call_analysis_metadata.phrase_matchers.id:"one_negative_custom_highlight" -latest_analysis.analysis_result.call_analysis_metadata.phrase_matchers.id:"multiple negative custom highlights"
Data source	Source of the conversation: Cloud Storage or Dialogflow	data_source.gcs_source: "*" data_source.dialogflow_source: "*"
Dialogflow conversation	Conversation that took place through Dialogflow, using a virtual agent	data_source.dialogflow_source.dialogflow_conversation="projects/123456789/locations/us-central1/conversations/some_dialogflow_conversation"
Dialogflow conversation profile ID	String for a unique Dialogflow conversation profile	dialogflow_conversation_profile_id="some_dialogflow_conversation_profile_id"
Duration	Length of time conversation lasted	duration<="20s"
Language	Spoken or typed language used in the conversation with language code in BCP-47 format and supported prefixes	language_code: "some_language_code"
Latest analysis create time	Time at which the latest analysis was created	latest_analysis.create_time>"1969-12-31T16:00:10-08:00" latest_analysis.create_time<$eq;"1969-12-31T16:00:20-08:00"
Latest summary	Match a portion of the most recent summary	latest_summary.text:"phrase to match" latest_summary.text:"and another" -latest_summary.text:"exclude this"
Sentiment score	Average user sentiment score for the conversation	latest_analysis.analysis_result.call_analysis_metadata.sentiments.sentiment_data.score>"0.1" latest_analysis.analysis_result.call_analysis_metadata.sentiments.sentiment_data.score<"0.2"
Silence duration	Length of the silence time the conversation has	latest_analysis.analysis_result.call_analysis_metadata.silence.silence_duration>"10s" latest_analysis.analysis_result.call_analysis_metadata.silence.silence_duration<$eq;"20s"
Smart highlight	Smart highlights to help flag certain phrases.	latest_analysis.analysis_result.call_analysis_metadata.intents.display_name:"one_positive_smart_highlight" latest_analysis.analysis_result.call_analysis_metadata.intents.display_name:"multiple positive smart highlights" -latest_analysis.analysis_result.call_analysis_metadata.intents.display_name:"one_negative_smart_highlight" -latest_analysis.analysis_result.call_analysis_metadata.intents.display_name:"multiple negative smart highlights"
Summary status	Whether a summary has been generated	Has a summary: latest_summary:"*" No summary: -latest_summary:"*"
Topic (also called Issue)	Topic or issue identified by Topic Modeling	latest_analysis.analysis_result.issues.issue="projects/123456789/locations/us-central1/issueModels/test_model/issues/issue_foo" latest_analysis.analysis_result.issue_model_id="projects/123456789/locations/us-central1/issueModels/test_model"
Transcript	Match text of the conversation	transcript.transcript_segments.text:"one_positive_term" transcript.transcript_segments.text:"multiple positive terms" -transcript.transcript_segments.text:"one_negative_term" -transcript.transcript_segments.text:"multiple negative terms"
Turn count	Number of conversation turns	turn_count>"10" turn_count<="20"

Operation filters

You can also filter conversations based on the information available from operations that have been performed on them. You can filter a conversation dataset based on the following operation attributes.

Filter term	Description	Reference code
Create time	Time at which the operation was created	metadata.create_time>"1969-12-31T16:00:10-08:00" metadata.create_time<="1969-12-31T16:00:20-08:00"
Conversation ID	String for a specific conversation	metadata.conversation_id="some_conversation_id"
Done	Final status of an operation	Completed: done="true" Not completed: done="false"
Operation Type	Type of operation performed on the dataset	Operation type is: metadata.operation_type="BulkAnalyzeConversations" Operation type is not: metadata.operation_type!="BulkAnalyzeConversations"
Status	Detailed status of the operation	Completed successfully: status_code="0" Failed but not aborted: status_code!="ABORTED"

To view the errors from running a bulk operation, you must use a curl command.

• Bulk analyze operations:

  curl -X GET \
   -H "Authorization: Bearer $(gcloud auth application-default print-access-token)" \
   -H "Content-Type: application/json; charset=utf-8" "https://contactcenterinsights.googleapis.com/v1/projects/$PROJECT/locations/$LOCATION/operations?filter=metadata.operation_type=\"BulkAnalyzeConversations\"&page_size=10"

• Bulk ingest operations (transcript or audio):

  curl -X GET \
   -H "Authorization: Bearer $(gcloud auth application-default print-access-token)" \
   -H "Content-Type: application/json; charset=utf-8"  "https://contactcenterinsights.googleapis.com/v1/projects/$PROJECT/locations/$LOCATION/operations?filter=metadata.operation_type=\"IngestConversations\"&page_size=10"