Overview

The Call Journey BI Connector Service is designed to enable users to more easily ingest EVS voice data into their own data environment.

The transcription engine handles speech to text conversion and provides additional metadata of each processed audio file including emotional intelligence, confidence scores, word start and end times, and other call information. This information is stored in the json file format in an organized fashion to provide the most content that can be derived from the audio source.

To enable the information stored within the json files to be more easily ingested into BI Applications (such as Excel, Power BI, Tableau, Amazon QuickSight and Emite), elements from the json file have been extracted and stored within a PostgreSQL database. Further, additional metrics have been calculated using the data within the json file. The database stores the information for one audio call in one row or record of the database.

The database contains three tables:

public generated_analytics_data
public generated_metrics_primary_data
public generated_metrics_secondary_data

Note: when a data value is not defined (null), the data name will not appear as an element in the JSON output, but it will be shown as null in the database.

CONTENTS

Backup

Importing and manipulating data may put your systems under load and could cause them to become inoperative. Therefore, backup all data before following any of the instructions in this Guide. TMA Comms Pty Ltd is not responsible for any loss of data.

Database Updates and Retention Periods

When you subscribe to the BI Connector service, data for the previous six months will be retrieved and loaded into the database. Thereafter, any json files created from new calls being processed will be added to the database. Any modified json files resulting from Wordbench apps being re-scored will be reloaded into the database, overwriting the previous record. Records will be kept in the database for 6 months from the date of upload. Therefore, for existing customers, up to 12 months of data may be in the database before records start to expire and be deleted from the database.

Note: the entire initial upload of six months data will be deleted from the database after six months.

Customers will be responsible for keeping permanent copies of the data after six months.

Updating Json Files

When modifying Wordbench apps, json files for calls already processed are not automatically updated. If you want the json file updated for these previous calls so that the revised app scores are reflected in the PostgreSQL database, please ensure that you select the box next to Update scores in file JSON:

Data Types, Rounding and Other Items

The tables below set out the names, type and description of the fields available in each of the tables within the database.

Note that all decimal fields are rounded to 2 decimal places.

Some fields may be highly correlated and therefore care needs to be taken when using these fields in any analysis (such as a regression analysis). For example:

The percentage of words spoken on the agent channel and client channel will sum to one (agentwordcountpercent + clientwordcountpercent =1). Therefore, correlation would be negative one (except for rounding differences).
It is likely that the number of interruptions on a call (count of the number of words both channels spoke over the top of the other channel) to be highly correlated to cross talk (total time in minutes each channel was speaking at the same time).

The key point here is to understand the data before using it in any analysis.

BI Connector Credentials & Data Security Requirements

The database is hosted on Amazon AWS. To ensure a secure and encrypted transmission of data you will require AWS certificates to be installed and enabled on the machine used to access the data. Details can be found on the AWS website:

https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/UsingWithRDS.SSL.html

See Appendix A on how to install the certificate on a Windows 10 local computer.

When you subscribe for the BI Connector, you will be provided with four parameters that will be required to connect and import data into your BI application:

Server or Host URL
Database name
Username
Password

Each BI application will have its own method for importing data.

Using Power BI with the BI Connector Service

Power Bi Dashboards have been created and made available via a Power BI template. Instructions for using the template can be found here: How to Use Power BI with the Call Journey BI Connector Service

Table Format Guide

Table 1: public generated_analytics_data

Field	Type	Description
agentchannel	Integer	The channel the agent is on.
agentclarity	Decimal	How clear the agent channel/speech is. Range is between 0 and 1, with 1 being the clearest.
agentei	String	Emotional intelligence (“ei”) for agent (see ei field).
agentgender	String	Agent gender prediction.
agentid	String	Reserved metadata field used to store the agent name or identifier according to the metadata provided by the client.
agentwords	String	List of words transcribed on the agent channel.
appdata	object	json object that stores metadata generated by Wordbench
audioproperties	String	Standard field using standard Genesys Cloud connection.
callback	String	Additional fields retrieved via the api from a Genesys Cloud connection (only retrieved if IVR redaction is enabled).
calldirection	String	Additional fields retrieved via the api from a Genesys Cloud connection (only retrieved if IVR redaction is enabled).
clientchannel	Integer	The channel the client is on.
clientclarity	Decimal	How clear the client channel/speech is. Range is between 0 and 1, with 1 being the clearest.
clientdata	object	json object that stores call metadata associated with the audio file. clientdata consists of one metadata object. This data is a copy of the json or xml input metadata file uploaded to Wordbench alongside the audio file. There is no fixed data name, and it is up to the user to define whether they want to pass any information into the output json files.
clientei	String	Emotional intelligence (“ei”) for client (see ei field).
clientgender	String	Client gender prediction.
clientwords	String	List of words transcribed on the client channel.
company	String	Corresponds to the Company field in Wordbench. Refer to Wordbench Management Guide which can be found in the Media Library on the website.
confidence	Decimal	A measure of how confident the speech recognition system is in its transcription results Range between 0 and 1 1 is most confident
conversationid	String	Standard field using standard Genesys Cloud connection. Important within the Genesys Cloud environment.
datetime	value	Reserved metadata field used to store the date and time that the call occurred according to the metadata provided by the client. For calls where no field is provided, this field will show the date and time of processing.
diarization	Decimal	This value provided in 2 speaker, 1 channel calls. Range between 0 and 1. 1 is best speaker separation. For dual channel calls where speakers are on separate channels, this field will not appear in the json file and will be shown as null in the database.
donedate	value	Date and time the file transcription was completed by the speech-to-text engine
duration	Decimal	Call duration in minutes
ei	String	Emotional intelligence consists of both acoustic and linguistic information. Events can be given the following values: Positive Mostly Positive Neutral Mostly Negative Negative
filename	String	EVS unique identifying. Will typically use the filename of the recorded call to enable the metrics to be matched to the original call.
folder	String	Corresponds to the Folder field in Wordbench. Refer to Wordbench Management Guide which can be found in the Media Library on the website.
insert_time	value	Date and time the database was populated.
insertid	String	Unique identifier to link the tables together within the database.
jm_version	String	Standard field using standard Genesys Cloud connection.
organisation	String	Corresponds to the Organization field in Wordbench. Refer to Wordbench Management Guide which can be found in the Media Library on the website.
original_json	json	Copy of json file as produced by the transcription engine.
overtalk	Decimal	Percentage of call when the agent talks over or interrupts the client. Equal to the number of turns where the agent initiated overtalk divided by the total number of agent turns.
queuename	String	Additional fields retrieved via the api from a Genesys Cloud connection (only retrieved if IVR redaction is enabled).
recordingid	String	Standard field using standard Genesys Cloud connection. Usually equal to the filename. Hence, less important.
requestid	String	Unique identifier generated by the transcription engine.
scorecard	object	json object that stores any application scores that have been calculated for the transcript. This object contains the Count and Coverage scores for each category in each application. For more information about how these scores are calculated and when they are included in a json output file, please refer to the Wordbench Application Development Guide.
sentiment	String	Linguistic sentiment value: Positive Mostly Positive Neutral Mostly Negative Negative Mixed (contains both Positive and Negative in the file).
silencepercent	Decimal	Percentage of overall duration that is silence. Equal to all non-speech time, calculated as call duration minus the sum of the duration of each word. If music and noise is not decoded to word-events, they would be counted as silence.
url	String	Location in Wordbench where the file can be viewed.
userid	String	Standard field using standard Genesys Cloud connection.
wordcount	Integer	Number of words in the transcription.

Table 2: public generated_metrics_primary_data

Field	Type	Description
agenteiscore (“ASat”)	Decimal	As for eiscore, but only for the agent channel, ranging between -1 (very negative) to +1 (very positive), with 0 (zero) being neutral. It can be used as a measure for the overall Agent Satisfaction (“ASat”) on the call by taking into account the words transcribed and the acoustic information. See eiscore for more information.
agentsentimentcountnegative	Integer	Count of the number of negative and mostly negative phrases made by the agent.
agentsentimentcountnet	Integer	agentsentimentcountpositive less agentsentimentcountnegative.
agentsentimentcountpositive	Integer	Count of the number of positive and mostly positive phrases made by the agent.
agentwordcount	Integer	Number of words in the transcript on the agent channel.
agentwordcountpercent	Integer	agentwordcount divided by wordcount.
clienteiscore (“CSat”)	Decimal	As for eiscore, but only for the client channel, ranging between -1 (very negative) to +1 (very positive), with 0 (zero) being neutral. It can be used as a measure for the overall Client Satisfaction (“CSat”) on the call by taking into account the words transcribed and the acoustic information. See eiscore for more information.
clientsentimentcountnegative	Integer	Count of the number of negative and mostly negative phrases made by the client.
clientsentimentcountnet	Integer	ClientSentimentCountPositive less ClientSentimentCountNegative.
clientsentimentcountpositive	Integer	Count of the number of positive and mostly positive phrases made by the client.
clientwordcount	Integer	Number of words in the transcript on the client client.
clientwordcountpercent	Integer	clientwordcount / wordcount
company	string	Corresponds to the Company field in Wordbench. Refer to Wordbench Management Guide which can be found in the Media Library on the website.
eiscore	Decimal	Each utterance is given an emotional intelligence descriptive value ranging from Positive to Negative. To translate these descriptive values into a number, each value is assigned a number as set out below. positive = +1 mostly_positive = +0.5 neutral = 0 mixed = 0 mostly_negative = -0.5 negative = -1 Each utterance is weighted by its duration (as a per cent of total utterance time). eiscore is the sum of these weighted values. Current settings will result in scores from -1 to +1 Note: eiscore = weighted sum of agenteiscore and clienteiscore. Alternative weighting conversion Some analysis may require only positive values. An alternative weighting could be from 0 to 1, such as: positive = +1 mostly_positive = +0.75 neutral = +0.5 mixed = +0.5 mostly_negative = +0.25 negative = 0 To convert the eiscore to a value using the above weightings, simply divide the value by 2 and add 0.5. That is: eiscore (alternative weighting) = (eiscore / 2) + 0.5
folder	string	Corresponds to the Folder field in Wordbench. Refer to Wordbench Management Guide which can be found in the Media Library on the website.
insert_time	value	Date and time the database was populated.
insertid	string	Unique identifier to link the tables together within the database.
organisation	string	Corresponds to the Organization field in Wordbench. Refer to Wordbench Management Guide which can be found in the Media Library on the website.
sentimentcountnegative	Integer	Count of the number of negative and mostly negative phrases made in the call on both channels. Note: sentimentcountnegative = agentsentimentcountnegative + clientsentimentcountnegative
sentimentcountnet	Integer	sentimentcountpositive less sentimentcountnegative
sentimentcountpositive	Integer	Count of the number of positive and mostly positive phrases made in the call on both channels. Note: sentimentcountpositive = agentsentimentcountpositive + clientsentimentcountpositive

Table 3: public generated_metrics_secondary_data

Field	Type	Description
agentcrosstalk	Decimal	Total time in minutes the agent was speaking while the client was speaking and the agent started speaking after the client.
agentcrosstalkpercent	Decimal	Percentage of total talk time that the agent was speaking while the client was speaking and the agent started speaking after the client. agentcrosstalkpercent = agentcrosstalk / talktime.
agentinterruptions	Integer	Count of the number of words the agent spoke while the client was already speaking.
agenttalkingspeed	Integer	Number of words per minute spoken by the agent. agenttalkingspeed = agentwordcount / agentutterancetime Care should be taken when interpreting this metric for low duration calls.
agenttalktime	Decimal	Total talk time in minutes for the agent. Sums the difference between the start and end times for each word spoken by the agent. Note: excludes the silence between each word in an utterance.
agentutterancetime	Decimal	Total utterance time in minutes for the agent. Sums the difference between the start and end times for each utterance for the agent.
clientcrosstalk	Decimal	Total time in minutes the client was speaking while the agent was speaking and the client started speaking after the agent.
clientcrosstalkpercent	Decimal	Percentage of total talk time that the client was speaking while the agent was speaking and the client started speaking after the agent. clientcrosstalkpercent = clientcrosstalk / talktime
clientinterruptions	Integer	Count of the number of words the client spoke while the agent was already speaking.
clienttalkingspeed	Integer	Number of words per minute spoken by the client. Note: based on utt_time which is derived at a utterance level and not a word level. clienttalkingspeed = clientwordcount / clientutterancetime Care should be taken when interpreting this metric for low duration calls.
clienttalktime	Decimal	Total talk time in minutes for the client. Sums the difference between the start and end times for each word spoken by the client. Note: excludes the silence between each word in an utterance.
clientutterancetime	Decimal	Total utterance time in minutes for the client. Sums the difference between the start and end times for each utterance for the client.
company	String	Corresponds to the Company field in Wordbench. Refer to Wordbench Management Guide which can be found in the Media Library on the website.
crosstalk	Decimal	Total time in minutes each channel was speaking at the same time. Derived at the word level by comparing the start and end time of each word to the start and end times or all other words. The sum of any overlap is crosstalk.
crosstalkpercent	Decimal	Percentage of total talk time that both channels were speaking. crosstalkpercent = crosstalk / talktime
folder	string	Corresponds to the Folder field in Wordbench. Refer to Wordbench Management Guide which can be found in the Media Library on the website.
insertime	value	Date and time the database was populated.
insertid	string	Unique identifier to link the tables together within the database.
interruptions	Integer	Count of the number of words both channels spoke over the top of the other channel.
organisation	string	Corresponds to the Organization field in Wordbench. Refer to Wordbench Management Guide which can be found in the Media Library on the website.
silence	Decimal	Silence on call in minutes. Equal to all non-speech time, calculated as call duration minus the sum of the duration of each word. If music and noise is not decoded to word-events, they would be counted as silence. silence = duration x silencepercent.
talktime	Decimal	Total talk time in minutes. Sums the difference between the start and end times for each word. Note: excludes the silence between each word in an utterance. talktime = agenttalktime + clienttalktime
utterancetime	Decimal	Total utterance time in minutes. Sums the difference between the start and end times for each utterance. utterancetime = agentutterancetime + clientutterancetime