Speech Analytics Configuration

NICE Uptivity Speech Analytics searches recorded calls from your contact center for specific words and phrases . You can specify the criteria it uses to determine which calls to analyze. You can also specify the words and phrases that it looks for using tags. Speech analytics comes with many prebuilt reports that you can use to gain insight into many aspects of your contact center. You can also customize your own reports.

Additionally, you can create classification rules to have speech analytics add tags to user fields in call records. Users can then use those tags to search for specific calls in the Web Portal. For example, you can create a tag for each marketing campaign, then create classification rules so that speech analytics inserts the right tag into the user fields of calls related to each campaign. Then your users can search in the Uptivity Web Portal for calls related to a specific marketing campaign.

For additional help in configuring Speech Analytics, see the Uptivity PDF guide Develop an Analytics Program.

Analytics Tags

Analytic tags determine the words and phrases that the speech search engine searches for in index files. The search process is continuous as long as new calls and index files are created. You can create new tags from scratch, or you can copy existing tags and modify them.

For each tag, you can specify a time period during which the speech engine actively searches calls for matching phrases. You can set the date range to search historical files, or you can set a start date but no end date. Without an end date, speech analytics searches all future calls for the tag. It will stop if you disable the tag or modify it to include an end date.

Performance

Speech engine performance depends primarily on the number of call audio hours processed and the NICE Uptivity Speech Analytics server’s CPU. The total number of tag phrases can also affect the speed of Speech Analytics processing. Customers with 3,000 or more phrases have reported slower processing speeds.

Analytic tags appear in the Tag Cloud quick filter in the Recorded Interactions list. If you create a large number of call tags, or if one particular tag occurs very frequently, the Tag Cloud may not display correctly. As a best practice, we recommend nightly indexing of the SQL tag data tables.

Reprocess Historical Speech Files

NICE Uptivity Speech Analytics can search for tags in historical speech files, also called phonetic access tracks (or PATs). PATs are index files of analyzed call interactions that Speech Analytics uses to identify tags. A PAT is considered historical if it was created any time before the day that a Speech Analytics tag is created.

The reprocessing job begins when you save the new analytics tag. When the new analytics tag is saved, the reprocessing job cannot be paused or stopped. Reprocessing jobs can recover from an interruption that may unintentionally stop the reprocessing job before the job is completed, such as a server reboot for example.

To monitor the status of a reprocessing job, you can monitor start and complete events in the Analytics Manager log file.

Reprocessing old files can delay the system processing new recordings. It's best practice to use analytics tags for historical reprocessing during your contact center's off-peak hours.

Considerations for Deleting Tags

Deleting tags is not recommended in most cases. Deleting a tag deletes all the phrase records generated using that tag. This can drastically affect reporting. Any reports that depend on historic information and comparisons will be unreliable if tags are deleted.

Speech analytics allows you to disable tags you don't want to be actively searched on. If the tag is needed sometimes but not always, it's better to enable and disable it as needed.

If you decide to delete tags, delete them when the Uptivity Speech Analytics system isn't processing many audio files (such as early in the morning). Tag deletion can cause performance issues for reporting and speech analytics processing. For example, if a tag is found 10,000 times, deleting it causes that number of records to be deleted from the database. This can require a lot of processing time and delay when the system begins processing new calls.

Tag Groups

You can create groups of tags that relate to each other. The Speech Category Summary and Speech Category Trending reports use tag groups to search for and organize data. Other reports use analytic tag groups to organize how tags appear on reports.

Deleting a tag group does not delete the individual tags in that group. However, if a report is based on a group, then that report will no longer get data.

Criteria Expressions

Criteria expressions determine which call audio files Speech Analytics should search for a given tag. When you create tags, you can assign a criteria expression to it. This means that Speech Analytics will only search for that tag in calls that meet the criteria in the selected expression.

If you don't assign a criteria expression to a tag, Speech Analytics searches all calls for the tag. This can result in a high volume of calls backing up in the system. For this reason, a minimum of one criteria expression is required.

You can create a single criteria expression and edit the criteria's text to select all needed call audio files. This approach can avoid the problem of having to check multiple expressions to see if the correct calls are being analyzed. Some users find it easier to manage multiple expressions based on language, client, or purpose. The number of expressions is one of many factors that affect indexing speed; there is no simple way to determine what effect the number may have on processing.

Criteria Selection

Selecting effective criteria is important. Poor call selection criteria means that Speech Analytics can miss some calls that should be analyzed, or analyze calls that are irrelevant. This can cause:

• Inaccurate (too low or too high) numbers reported for compliance, process improvement, and agent assessment.
• Missed opportunities for improvement or discovery.
• Unnecessarily slow indexing and searching of calls and generation of reports.

Call selection criteria and the number of criteria expressions depends on how agents are organized and how calls are directed over your telephony network. For example, all Mexican Spanish calls may be received via one queue or telephone number, or calls for a client that requires a script may be directed to specific agents.

Call selection criteria and naming can be based on:

• Language.
• Client (if your organization has multiple internal or external clients, some may want information tracked or monitored while others do not).
• Purpose or function such as compliance, security, or quality training.

Classification Rules

Analytic classification rules add speech tags to user fields in call recording records. Users can then search for these call records in the NICE Uptivity Web Portal. You can create rules to classify calls in any way that's meaningful to your organization. For example, by product or service, quality of customer service, compliance, or marketing campaign.

Classification rules:

• Can only be applied to new calls; calls that existed prior to a rule's creation cannot be classified.
• Do not indicate when a tag occurred in a call.
• Do not affect Tag Cloud quick filter.
• Are not used in pre-built reports.
• Can be used in ad hoc call recording reports.

Considerations

You can use analytic classification rules to insert, replace, or append data to any of the custom user data fields, the ACD Group field, or the ACD Gate field. Consider the following when deciding what action your rule should take:

• User data fields have varying size limits. For example, the User1 field is limited to 20 characters. When you create a rule, NICE Uptivity will warn you if your data exceeds the field limit. However, if the rule could append data that would exceed the limit, NICE Uptivity will not append that new data.
• Be sure that the field used in your rule is not already being used for another purpose, especially if you choose to replace existing data. The fields that Speech Analytics can use can also be updated by other means. The custom user fields can be populated by Legacy Desktop Analytics (Fusion), NICE Uptivity On-Demand, an API call to another application, or manually in the Web Portal. The ACD Gate and ACD Group fields are typically populated via your telephony integration.