An interaction is assigned to a regular category if it meets the conditions that are defined for inclusion in the category. The conditions can include a wide variety of interaction characteristics, including the program the interaction belongs to, topics that were recognized in it, non-linguistic events that were detected in it (calls only), its metadata properties, and even other categories to which it belongs.
Define your category by including conditions that will recognize interaction characteristics associated with the specific category's business issue.
To configure a category condition
Configure one or more of the following conditions:
If you are adding a condition to an existing expression, select the element in the expression before which you want to insert the new element.
Metadata is information about an interaction that is gathered by the external recording systems and relayed to SpeechMiner together with the interaction's data.
For example, interaction metadata can include the agent's name, the customer's location, the interaction duration, and so on.
Metadata category conditions are based on the metadata values available in your system. The conditions you define can be as simple as one metadata value or they can be complex expressions that specify the relationships among a number of types of metadata. Consider, for example, a sales campaign that is undertaken in Wyoming and in Texas. Certain agents in each location target senior citizens, while others target families with teenage children. To assign senior-citizen interactions to a Category, you could define a condition that specifies either Agent1 or Agent2 if the customer is in Wyoming and Agent3 if the customer is in Texas.
Drag Meta Data to the area below the conditions bar.
The fields required to configure the element appear below the Type parameter.
In the first field on the left, specify the type of metadata in one of two ways:
If the type you want appears in the drop-down list, select it. For example, select Workgroup\Agent if you want to define a condition based on the workgroup or agent that handled the interaction.
Manually enter the type in the field. For example, if your system gathers data about an interaction's direction under the name interactiondirection, enter interactiondirection.
In the remaining fields, specify the condition. In some cases, you can select an operator (<, >, or =) in the second field and specify a value in the third. In other cases, you can only specify a value (and the = operator is assumed). You can specify a value by selecting it from the list or by typing it manually into the text field on the right.
Some of the value options may be arranged hierarchically. For example, if you select Workgroup\Agent, a list of workgroups is displayed. You can select a workgroup, or click the + beside the name of a workgroup to expand the list and display the names of the agents in the specific workgroup. If you select a workgroup, all members of the workgroup are included in the condition; if you expand a workgroup and select an agent, only that agent is included.
If you want the condition to include all interactions except those that match the condition, select Not found.
Non-linguistic events include various sounds that SpeechMiner detects in an interaction, such as busy tones, key presses, and music or noise. Depending on the program settings of the interaction, non-verbal indications of agitation may also be detected. In addition, the non-linguistic event called Cross Talk is identified when two speakers are talking at the same time. This type of event can only be identified in systems in which dual-channel recording is employed.
In some cases, an element can include more than one type of related event. For example, you can select Silence, Busy Tone, and Dial Tone together. When you select multiple types of non-linguistic events, if any of those events is found in the interaction, it meets the condition.
Each non-linguistic event that is identified by SpeechMiner has a start time, an end time, and a type. For example, if SpeechMiner identifies silence in an interaction, this is a non-linguistic event whose start time is the beginning of the silent period, whose end time is the end of the silent period, and whose type is "Silence."
You cannot select Agitation along with other types of events.
Drag Non Linguistic to the area below the conditions bar.
Select one or more of the following types of non-linguistic events:
Types of non-linguistic events that cannot be selected along with the selected type, become inactive.
Dial Tone: A dial tone.
Ringback: A signal used in PSTN (Public Switched Telephone Networks - standard "land lines") to indicate that the line is being called or an incoming interaction is present.
Silence: Silence, usually indicating that the Mute button was held down on one of the handsets.
Busy Tone: A busy signal.
Music and Noise: Music or noise, generally indicating the interaction was on hold.
DTMF: Key press on a touch-tone phone. 16 different keys can be identified using DTMF (Dual Tone Multi Frequency). Select any to accept any key press.
Agitation: Non-verbal expressions of frustration and anger, such as deep sighs, grunts or rapid changes in pitch.
Cross Talk: When the agent and customer talk at the same time.
If you selected Silence, Busy Tone, Music/Noise, Dial Tone, Ring Back Tone, or Cross Talk, define the amount of time that the selected types of non-linguistic events must occupy in the interaction total as follows:
Field
Description
From / To
Define the minimum amount of time that the selected types of non-linguistic events must occupy in the total interaction time in one of the following ways:
Seconds: The minimum and maximum number of seconds from the beginning of the interaction at which the event types must be found and end.
Percentage: The minimum and maximum proportion of the interaction (percentage), at which the event type must start and end. (Insert the % character after the value to indicate it is a percentage.)
Accumulated Amount
The values in the From and To fields refer to the total amount of time in the interaction audio in which the selected types of events were found, even if the total consists of more than one distinct event.
Continuous
The values in the From and To fields refer to individual events in the interaction audio. That is, an event only meets the condition if it lasted at least as long as the value in the From field and not longer than the value in the To field.
Configure the following Position settings:
Field
Description
Not found
Select this option if you want the condition to include all interactions except those in which the selected types of non-linguistic events were found.
From
Define the earliest position in the interaction audio at which the non-linguistic event can be found and select one of the following types of reference points:
Secs Abs: The number of seconds from the beginning of the interaction.
Secs Abs from End: The number of seconds from the end of the interaction. That is, the event must be found no earlier in the interaction than this number of seconds from the end of the interaction.
Secs Rel to Start of Prev: The number of seconds relative to the beginning of the previous item in the expression. This option will work only if the operator connecting the two elements is Conditional AND (->). Otherwise, selecting this is the same as selecting Secs Abs.
Secs Rel to End of Prev: The number of seconds relative to the end of the preceding item. This option will work only if the operator connecting the two elements is Conditional AND (->). Otherwise, selecting this is the same as selecting Secs Abs from End.
To
Define the latest position in the interaction audio at which the non-linguistic even may be found.
A topic element specifies that a topic must be found (or must not be found) by SpeechMiner in an interaction for the interaction to be assigned to the specific category.
A program element specifies which interactions belong to the program for the interaction to be assigned to the specific category.
Drag Topic to the area below the conditions bar.
Select the topic to which the interactions should belong.
Configure the following settings as necessary:
Condition
Description
Not found
Select this option if you want the condition to include all interactions except those in which the specific topic was found.
Occurrence
The number of occurrences indicates the minimum number of times a topic must be found for the condition to be met. If the minimum is not met, SpeechMiner proceeds as if the Topic was not found.
The Strictness level indicates the minimum strictness level required for SpeechMiner to recognize the Topic. If the minimum Strictness level is not found, SpeechMiner proceeds as if the topic was not found.
The Confidence level (a range from 1 to 100), indicates the minimum level of confidence for the recognition of the phrase. If the SpeechMiner Confidence Level does not meet this minimum, SpeechMiner proceeds as if the phrase was not found.
Position
The Secs options should be selected to match audio interactions and the Words options should be selected to match text interactions. If you define 0 seconds from the start and 0 seconds from the end of the interaction, the entire (audio or text) interaction will be analyzed.
From: Define the earliest position in the interaction at which the topic should be found and then select one of the following.
Secs Abs: The number of seconds from the beginning of the interaction.
Secs Abs from End: The number of seconds from the end of the interaction. That is, the event must be found no earlier in the interaction than this number of seconds from the end of the interaction.
Secs Rel to Start of Prev: The number of seconds relative to the beginning of the previous item in the expression. This option will work only if the operator connecting the two elements is Conditional AND (->). Otherwise, selecting this is the same as selecting Secs Abs.
Secs Rel to End of Prev: The number of seconds relative to the end of the preceding item. This option will work only if the operator connecting the two elements is Conditional AND (->). Otherwise, selecting this is the same as selecting Secs Abs from End.
Words Abs: The number of words from the beginning of the interaction.
Words Abs from End: The number of words from the end of the interaction. That is, the event must be found no earlier in the interaction than this number of words from the end of the interaction.
Words Rel to Start of Prev: The number of words relative to the beginning of the previous item in the expression. This option will work only if the operator connecting the two elements is Conditional AND (->). Otherwise, selecting this is the same as selecting Words Abs.
Words Rel to End of Prev: The number of words relative to the end of the preceding item. This option will work only if the operator connecting the two elements is Conditional AND (->). Otherwise, selecting this is the same as selecting Words Abs from End.
To: Define the latest position in the interaction at which the preceding Topic should be found and then select one of the available options.
Speaker
Select a speaker type to instruct SpeechMiner to find and analyze Topics where only the Agent, Customer or both speak.
Click OK.
Important
The Position > Words options are only available when Text Analytics is enabled.
A category element specifies that an interaction must (or must not) belong to another category for the interaction to be assigned to the category that is currently being configured.
Although you can create circular category references, the categories in which they appear cannot be applied to SpeechMiner. For example, if category A is an element of the expression that defines category B, category B cannot also be an element of the expression that defines category A. If they are, the circular references are indicated in the Apply dialog box, in the Message column of each Category. For additional information about applying changes to SpeechMiner, see Applying Changes.
Important
Including a Category element in the expression that defines a category is not the same as changing a category into a sub-category of the other. For additional information about sub-categories, see About Categories.
Drag Category to the area below the conditions bar.
From the Category list, select the category the interaction must belong to.
If you want the condition to include all interactions except those that belong to the specific category, select Not found.
A Speaker-Type element specifies that either an agent or customer or both must have taken part in an interaction for the interaction to be assigned to the category. This information is gathered by the recording system and passed on to SpeechMiner along with the interaction data. The exact type of information available depends on the recording system that is employed in your company.
Some recording systems also indicate the times in the interaction when each speaker was heard. For this reason, you can specify the times in the interaction when the type of speaker must have taken part in the interaction for the interaction to meet the category condition. For example, you could specify that a supervisor must have taken part in the last 120 seconds of the interaction.
Drag Speaker Type to the area below the conditions bar.
From the Speaker Type list, select the type of speaker for which SpeechMiner should search in association with the specific category.
If you want the condition to include all interactions except those that belong to the specific category, select Not found.
Configure the following settings as necessary:
Condition
Description
Position
From: Define the earliest position in the interaction at which the speaker should be found and then select one of the following:
Secs Abs: The number of seconds from the beginning of the interaction.
Secs Abs from End: The number of seconds from the end of the interaction. That is, the event must be found no earlier in the interaction than this number of seconds from the end of the interaction.
Secs Rel to Start of Prev: The number of seconds relative to the beginning of the previous item in the expression. This option will work only if the operator connecting the two elements is Conditional AND (->). Otherwise, selecting this is the same as selecting Secs Abs.
Secs Rel to End of Prev: The number of seconds relative to the end of the preceding item. This option will work only if the operator connecting the two elements is Conditional AND (->). Otherwise, selecting this is the same as selecting Secs Abs from End.
To: Define the latest position in the interaction at which the preceding Topic should be found and then select one of the available options.
A Text element for audio interactions specifies that a phrase containing a sequence of one or more whole words must be found by SpeechMiner within the transcript of an interaction, for the interaction to be assigned to the category. Also, the words should be separated by single spaces and additional symbols and punctuation should not be included in the phrase. A Text element for text interactions (such as, email, chat, and so on), can contain additional symbols.
To meet the category condition, the phrase must be found exactly as it is entered in the Text element definition (not case sensitive). For example, "i want to repay my loan" and "I want to Repay My loan" are treated as identical phrases.
Drag Text to the area below the conditions bar.
In the Text field, enter the phrase.
The phrase should consist of a sequence of one or more whole words. The words should be separated by single spaces. No punctuation or additional symbols should be included in the sequence.
Condition
Description
Not found
Select this option if you want the condition to include all interactions except those that include the phrase.
Occurrence
The number of occurrences indicates the minimum number of times the phrase must be found for the condition to be met.
The Strictness level indicates the minimum strictness level required for SpeechMiner to recognize the phrase.
The Confidence level (a range from 1 to 100), indicates the minimum level of confidence for the recognition of the phrase. If the SpeechMiner Confidence Level does not meet this minimum, SpeechMiner proceeds as if the phrase was not found.
Position
The Secs options should be selected to match audio interactions and the Words options should be selected to match text interactions. If you define 0 seconds from the start and 0 seconds from the end of the interaction, the entire (audio or text) interaction will be analyzed.
From: Define the earliest position in the interaction at which the topic should be found and then select one of the following.
Secs Abs: The number of seconds from the beginning of the interaction.
Secs Abs from End: The number of seconds from the end of the interaction. That is, the event must be found no earlier in the interaction than this number of seconds from the end of the interaction.
Secs Rel to Start of Prev: The number of seconds relative to the beginning of the previous item in the expression. This option will work only if the operator connecting the two elements is Conditional AND (->). Otherwise, selecting this is the same as selecting Secs Abs.
Secs Rel to End of Prev: The number of seconds relative to the end of the preceding item. This option will work only if the operator connecting the two elements is Conditional AND (->). Otherwise, selecting this is the same as selecting Secs Abs from End.
Words Abs: The number of words from the beginning of the interaction.
Words Abs from End: The number of words from the end of the interaction. That is, the event must be found no earlier in the interaction than this number of words from the end of the interaction.
Words Rel to Start of Prev: The number of words relative to the beginning of the previous item in the expression. This option will work only if the operator connecting the two elements is Conditional AND (->). Otherwise, selecting this is the same as selecting Words Abs.
Words Rel to End of Prev: The number of words relative to the end of the preceding item. This option will work only if the operator connecting the two elements is Conditional AND (->). Otherwise, selecting this is the same as selecting Words Abs from End.
To: Define the latest position in the interaction at which the preceding Topic should be found and then select one of the available options.
Speaker
Select a speaker type to instruct SpeechMiner to find and analyze Topics where only the Agent, Customer or both speak.
Click OK.
Important
The Position > Words options are only available when Text Analytics is enabled.
