Rss Categories

CPA Tuning

Reference Number: AA-02464 Views: 6645 0 Rating/ Voters

Tuning interactions is an important aspect of understanding and improving your application's performance. The same is true when working with Call Progress Analysis (CPA)

The Speech Tuner is the perfect tool to help you get a much deeper understanding of how your application is performing, and it also includes specific functionality to handle CPA interactions

After gathering data from your application, you can load this into the Speech Tuner and begin tuning, which involves Transcribing your data, and running Tests.

CPA Statistics

The functionality of CPA and the timing related to classifications of HUMAN RESIDENCE and HUMAN BUSINESS versus UNKNOWN SPEECH for longer utterances were calculated by LumenVox engineers over a long period of time, using a large amount of statistical data to perform an analysis.

This is where the default values for CPA_HUMAN_RESIDENCE_TIME (1800 ms) and CPA_HUMAN_BUSINESS_TIME (3000 ms) came from.

As you can see from the above illustration, we calculated the statistical average for HUMAN RESIDENCE is below 1800 ms, HUMAN BUSINESS to be between 1800 ms and at most 3000 ms, and anything longer than that is most likely a recorded message of some sort, so is classified as UNKNOWN SPEECH. The decision as to how to handle UNKNOWN SPEECH classification is very much implementation dependent, so it is up to your application to determine how to respond to these classifications.

An additional classification is also possible, where no human speech was detected at all in the audio stream. This is classified and returned in the results as UNKNOWN SILENCE. Again, how to react to this classification is application-specific, however most treat this as if a machine had answered (similar to UNKNOWN SPEECH)

The values for these settings, as described in our Grammars in CPA and AMD article can be changed by the application when needed. Perhaps your user population is mostly using mobile phones, or perhaps your users speak different languages, which causes some variance in the timings. Whatever the reason, it is best to gather information and look at these statistically using as much data as possible.

It is also worth checking the statistics to make sure your other application logs match what the Speech Tuner indicates. A significant difference might indicate some issue in the system or call flow somewhere.

Understanding CPA Classification

Although it was possible to determine the cause of a particular CPA classification when reviewing logs and settings in use when a CPA interaction was processed, starting with version 19.1 of the Speech Tuner, a new option is available to users which very clearly describes the logic behind why a certain CPA classification was made.

This new classification information is available in the Interactions Properties dialog window, which can be accessed by right-clicking on an interaction and selecting the Properties option from the menu that appears.

CPA Analysis

On the Answers tab of the dialog, in addition to detailed information about the result that was provided (which is used to display some of the information in the Interactions List), there is a new Call Progress Analysis Summary section below, which details how the CPA classification was calculated, based on the CPA settings that were in effect at the time.

As you can see from the example above, there was 420 ms of silence at the start of the audio stream, followed by human speech that lasted for around 2220 ms.  This length of speech (2220 ms) is greater than the time specified for HUMAN RESIDENCE, which was set to the default value of 1800 ms and less than the time specified for the limit of HUMAN BUSINESS, which was set to the default value of 3000 ms. This is a clear indication that the audio and human speech within was clearly within the range of HUMAN BUSINESS, which is what was returned.

Similar summary information for other CPA classification to help users fully understand the logic behind those classifications.

In addition to this CPA Analysis, an additional check is also made to determine if any potential issues were detected with the CPA settings, which will be reported immediately below the CPA Analysis Summary shown above. Any issues detected are described in some detail along with suggestions where applicable, however these are not errors, but warnings about using conflicting or inappropriate settings that can lead to unexpected application behavior.