Using Advanced Search with Interaction Analytics

The “Advanced search” feature in the Talkdesk Interaction Analytics™ application provides the ability to uncover deeper cases, as this option allows for more complex queries to be performed. 

 

In this article, you will find information on:

 

Accessing the Advanced Search feature

1-2..png

When accessing the “Search” page within the Interaction Analytics application, select the “Advanced search” toggle [1]

Tip: Check the query hint [2] displayed whenever the “Advanced search” toggle is enabled. This way you can double-check the Technical Name (e.g., interaction_started) of a given field, as an example of how to use it in a custom query. Note: To learn more about each filter, please check this article.

 

Editing Queries

Editing_queries_1.png

Users will be able to edit queries by switching the “Advanced Search” toggle to the right [1], whenever required.

 

Query examples

query_examples_1.png

Below, you can see several query examples that might be helpful while editing/typing a keyword or phrase in the “Query editing” search bar [1]

 

Logical Operators

Filter by a conjunction (AND) operator example:

  • All utterances that contain both the “hi” and the “hello” keywords:
("hello" AND "hi")

 

Filter by disjunction (OR) operator examples:

  • All utterances where the sentiment has been labeled as “positive” or “negative”: 
sentiment_label:("POSITIVE" OR "NEGATIVE")
  • All utterances labeled with the intent “Order returns” or “Order information”:
intent_value:("Order returns" "Order Information")

Note: query intent_value:("Order returns" OR "Order Information") would produce the same result.

 

Filter by negation (NOT) operator example:

  • All utterances labeled with the intent “Satisfaction” but don't contain “virtual agent” keyword
NOT message_text:(“virtual agent”) AND intent_value:("Satisfaction")

Filter by complex boolean operands' logics example:

  • All utterances that either contain one of the “hi” or “hello” keywords and were matched with intent “Order Information” or were matched with the “Order returns” intent and have a positive sentiment label:
(message_text:("hi" OR "hello") AND intent_value:("Order Information")) OR (intent_value:("Order returns") AND sentiment_label:("POSITIVE"))

 

Comparison Operators

Filter by equal to operator example:

  • All utterances where the sentiment has been labeled as “positive”:
sentiment_label:("POSITIVE")
  • All utterances that contain the “hello” keyword:
("hello")

Filter by greater than operator example:

  • All utterances that were matched with an intent with a confidence score of at least 0.5:
intent_confidence:[0.5 TO *]

Filter by less than operator example:

  • All utterances that were matched against an intent with a confidence score of up to 0.8:
intent_confidence:[* TO 0.8]

Filter by between range operator examples:

  • All utterances that were matched against an intent with confidence between 0.5 and 0.8:
intent_confidence:[0.5 TO 0.8]
  • All utterances that were matched against an intent with a confidence between 0.5 and 0.8, not including the 0.8 value:

intent_confidence:[0.5 TO 0.8]
  • All utterances that were matched against an intent with confidence between 0.5 and 0.8, not including the 0.5 value:
intent_confidence:{0.5 TO 0.8]

 

Proximity operators

Filter by one occurrence of any character example:

  • All utterances with keywords that start with “tes” followed by one more character (e.g., return results like “test”):
(tes?)

Filter by any sequence of characters examples:

  • All utterances with keywords that start with “tes” followed by any given number of characters (e.g., return results like “test”, “tests”, “testing”, among others): 
(tes*)

Note: Since the search is case-sensitive, you must specify lower and/or upper cases in the query search, so that the output of the record meets the desired criteria. If you need to search for keywords or phrases of a specific agent, you must include the exact username in your search. E.g. There is an agent named “John Doe”. To retrieve the correct results, you would need to type:

message_text:* AND message_agent_id:(John Doe)
  • Or, if you need more than one agent named John, you can use the search for any sequence of characters.
message_text:* AND message_agent_id:(John*)

Filter by fuzziness example:

  • All utterances that contain both “hello” and “today” keywords separated by up to 5 other keywords: 
("hello today"~5)

 

Dealing with different data types

Filter by dates examples:

  • All utterances for interactions starting from 2021-11-08T00:00:00: 
interaction_started:[2021-11-08T00:00:00 TO *]
  • All utterances for interactions started before 2021-11-09T23:59:59:
interaction_started:[* TO 2021-11-09T23:59:59]
  • All utterances for interactions started between 2021-11-08T00:00:00 and 2021-11-09T23:59:59:
interaction_started:[2021-11-08T00:00:00 TO 2021-11-09T23:59:59]

Filter by number examples:

AND intent_confidence:[0.5 TO 0.7]
AND intent_confidence:[0.88 TO 0.99]

Filter by special fields examples:

interaction_started:[2022-11-02T00:00:00 TO 2022-11-04T23:59:59]

Filter by channels queries examples:

  • Voice calls
AND (source_system:STT AND message_channel:VOICE)
  • Autopilot Digital
AND (source_system:VA AND message_channel:WEBCHAT)
  • Autopilot Voice
AND (source_system:VA AND message_channel:VOICE)

Querying with multiple filters:

  • Filtering outbound voice calls and showing only utterances spoken by the agent with the intent_id 298255eb-7f64-4ab2-b0c7-5353e5af713c in the last 3 months.
message_text:* AND interaction_started:[2022-08-09T23:00:00.000Z TO 2022-11-10T23:59:59.999Z] AND ((source_system:STT AND message_channel:VOICE)) AND message_direction:("OUT") AND message_participant:("AGENT") AND intent_id:("298255eb-7f64-4ab2-b0c7-5353e5af713c") AND (source_system:STT OR source_system:VA)
  • Filtering Autopilot Voice calls and showing only Negative sentiment spoken in en-US language.
message_text:* AND stt_language_code:("en-us") AND ((source_system:VA AND message_channel:VOICE)) AND sentiment_label:("NEGATIVE") AND (source_system:STT OR source_system:VA)

 

Filters Mapping

To build a query, you must use the Technical Names listed below:

Filter Name

Technical Name

Data type

“When”

interaction_started

Timestamp

“Channels”

source_system AND message_channel*

Text

“Direction”

message_direction

Text

“Ring Groups”

ring_groups

Text

“Speakers”

message_participant

Text

“Agents”

message_agent_id

Text

“Interaction ID”

interaction_id

Text

“Language”

stt_language_code

Text

“Intents”

intent_value

Text

“Intent Threshold”

intent_above_threshold

number

“Sentiment”

sentiment_label

Text

All Articles ""
Please sign in to submit a request.