System Setup

An Administrator can configure App Profile Settings, Ghost Call, CTI Language Engine Mapping, Entity Catalog Controller and Redaction.

App Profile Settings

System provides the ability to switch on or off the following features for a client/tenant.

ACW - Allows Admin to turn on/off this feature 'ACW'. By default, this option is turned on. If this feature 'ACW' is turned off, Agent cannot receive automated call summarization and disposition output from the application.
Alerts - Allows Admin to turn on/off this feature 'Alerts'. By default, this option is turned on. If this feature 'Alerts' is turned off, Agent and Supervisor cannot receive the real time Alerts on Agent desktop and Supervisor desktop, respectively.
In Call - Allows Admin to turn on/off this feature 'In Call'. By default, this option is turned off. If this feature is turned on, then the NLP engine is configured to provide real time results in order for intents and entities to get detected midway through a turn to get better response time for U-Assist In-Call use cases.
Limitation: It is recommended that the "In Call" feature be left turned off to allow the NLP Engine to work in real time.
Promises - Allows Admin to turn on/off this feature 'Assurance'. By default, this option is turned on. If this feature 'Assurance' is turned off, Agent cannot view the promises being automatically detected during the course of a live conversation with the customer.
Speech based alerts - Allows Admin to turn on/off this feature 'Speech based alerts'. By default, this option is turned on. If this feature 'Speech based alerts' is turned off, Agent and Supervisor cannot receive the emotion alerts on Agent desktop and Supervisor desktop, respectively.

Ghost Call Configuration

In a contact center, there could be calls where the voice of the agent/customer/both is not clear or calls that are too short possibly because of poor network. Agents need not spend their time for ACW on these types of calls, as there will not be any meaningful conversation on these calls.

U-Assist has the capability to identify and ignore contacts based on silence in agent/customer/both channels for specific configurable durations. These contacts will not have alerts/summaries/disposition generated and are referred to as “ghost” calls or “cancelled” calls. However, for certain kinds of cancelled calls, the call will be allowed to go on, though the alerts/summaries/dispositions will not be generated.

Note

The cancelled/ghost calls do not appear in the feedback loop.

Filter Conditions

This table provides information on the conditions that can be configured to filter the contacts. The error codes are shown in the contact detailed report.

Condition	Error Code	Message
If the first ‘x’ seconds audio is silence in both the channels (Agent and Customer)	1201	silent
If the entire call has silence in one of the channels (either Agent or Customer)	1202	silent agent
	1203	silent customer
If the duration of call is less than ‘x’ seconds	1204	too short
If the duration of call is between ‘y’ and ‘z’ seconds	1205	potential reject

Note

Silence detection is only applicable for the audio stream and not applicable for the text stream input of the contact.

The first condition can be validated during the course of the call and mark as ignored once the condition is met. All the other conditions can only be determined when the application receives the end event.

Note

If any of the above conditions are met, contact can be marked as ignored and further processing of the alerts and summary is ignored.

Configuration of filtering criteria

Administrator can configure various parameters for ghost call identification.

Parameter Name	Description
Enable	On – Indicates Ghost contact filtering feature is enabled in the U-Assist Application Off - Indicates Ghost contact filtering feature is not enabled
Minimum Contact Duration (milliseconds)	It refers to calls shorter than a preconfigured threshold. Contact duration must be greater than or equal to this minimum value.
Maximum Silence Duration (milliseconds)	It refers to silence on both channels (agent and customer) for a preconfigured threshold or silence on any one channel (agent or customer) for a continuous (preconfigured) period. Silence duration must be lesser than or equal to this maximum value.
Potential Reject Threshold (milliseconds)	These are calls that are longer than the Minimum Contact Duration but shorter than another preconfigured duration (referred to as Potential Reject Threshold). On the agent desktop, for all such calls, the agent will have the option to either submit the ACW or entirely reject the ACW. For example, calls between 10 - 45 seconds may have a valid conversation, but in most cases, calls get disconnected within 45 seconds because there was a challenge in the connection and hence the conversation needs to be aborted.

CTI Language Mapping

U-Assist application passes audio chunks and receives the transcripts for the audio chunk from the corresponding ASR Engines. This feature allows an Administrator to select the Organization and Business Process, as well as the language to be used for call transcription. Administrator can also mention a Skill Code which is mapped to the ASR Engine in the Cloud Admin Portal.

For more information on ASR Engine Configuration, click here.ASR Configuration in Cloud Admin Portal

The following table details the CTI and ISO language codes of different languages supported by the U-Assist.

CTI Language Code	ISO Language Code	Language
EUN	en-us	English-USA
SUN	es-us	Spanish-United States
SMN	es-mx	Spanish-Mexican
JAN	ja	Japanese
EUK	en-us	English-USA
JAK	ja	Japanese
EUU	en-us	English-USA
EIU	en-in	English-Indian
EPU	fil-en	English-PH
ARU	ar	Arabic
EGU	en-gb	English-Great Britain
EIU	en-in	English-India
GEU	de	German
HIU	hi-en-in	Hinglish-India
SUU	es-us	Spanish-United States
SMU	es-mx	Spanish-Mexico
JAU	ja	Japanese
EAU	en-au	English-Australia

Add CTI Mapping

Click Add CTI Mapping to add a new CTI language mapping. A new row is added in the CTI Language Mapping section.
Select Organization from the drop-down list.
Select Business Process from the drop-down list. The Business Processes created for the selected organization will be listed in the drop-down list.
Mention the CTI language code and select the corresponding ISO language code in the respective fields.
Mention the Skill Code for the selected Business Process. The same skill code must be mapped to the ASR Engine configuration.
Click Save and Close to save the CTI language mapping details.
Click on the ellipsis () icon to edit or remove the selected CTI language mapping.

Entity Catalog Controller

It is a repository of Entities that are required for a tenant. Once the entity catalog is used for training AI entities, it can be imported into the Analyst Entity Listing page. When NLP entities are created by the analyst, they need to be mapped to one of the entities in the catalog.

Important

When mapping NLP entities to the catalog, the NLP entity type should be the same as the entity type in the catalog.

In this way, a subset of the annotated calls can be used for testing accuracy of AI entities or NLP entities. The entity catalog is not relevant for rule entities or complex entities.

All annotation work is based on this entity catalog, and hence the catalog should ideally not change for a given deployment unless there is a change request that is documented.

Note

Any changes made to the catalog will have an impact to annotation work done on the previous version of the catalog. Re-annotation work will be necessary for all changes made to the catalog, whether they be creations, updates, or deletions.

An Administrator can create, update, delete and view Entity Catalogs which are listed in the Analyst Entity Listing page.

Parameter Name	Description
Id	Catalog ID. This value is automatically generated during entity catalog creation.
Name	Entity catalog name
Type	Entity type

Warning

Please ensure that at least one catalog is created for each AI entity with same name. If you are editing any catalog which is associated with an AI entity, edit with caution. Do not edit unless the AI entity name is changed.

Create Entity Catalog

Click Add Entity to add a new entity catalog. A new row is added in the Entity Catalog Controller section.
Mention the name of Entity Catalog.
Select entity type from the drop-down list. U-Assist provides built-in System Entities that can match with a wide range of commonly-used data such as date, time, email address, and so on. For more information on built-in Entity type, click here.
Click Save and Close to save the entity catalog details.
Click on the ellipsis () icon to edit the entity catalog details or remove the selected row.

Built-in Entity Types

Given below are the list of built-in System Entities:

Entity Type	Description
Alphanumeric	Entities which contain alphanumeric characters, i.e., numbers 0 to 9 and/or characters A to Z. Example: X33456, M B H J 1106, E 66501, etc. Alphanumeric Entity type is typically linked to the Entities, which include CVV, credit card number, credentials, account number, password, pin, product reference number, ticket number, device identification number, etc.
Boolean	`Entity which returns a Boolean value. (Affirmation based on the agent or customer’s colloquial response to a question – for example, yes/yup/yeah or no/nope response.) Best Practice: More than other entities, this entity is more dependent on the configured keyphrases. So please pay attention to the keyphrases, as we rely on negation words to return True or False, which may not be predictable for other languages.
Complex	Complex entities use REST APIs to retrieve data from various data sources like external CRM systems, other databases, and external APIs.
Date	Entity which contains any date or date range. Example: 11/20/2020, 05/20/20 to 05/25/20. Best Practice: Simple relative dates such as today, yesterday, last month, etc., are also covered. Recommend to customers about training agents in such a way that dates are always spoken by agents in a prescribed structure for good accuracy in model detection.
Date Time Duration	Entity which contains a combination of date, time, and duration, or any three of these Entity types individually. Example: tomorrow \| 5:00 p.m, 3/3/2024 , 8/25/ \| 4:00 p.m, yesterday \| 6:00 p.m
Duration	Entity which contains a duration in time, days, weeks, months or years. Example: 7 days, 5 business days, 2 to 3 weeks/months/years.
Email	Entity which defines any email id(s). Best Practice: Do not use this entity type unless you absolutely need to. Please contact Uniphore Support team before using.
Float	Entity which contains any floating-point numbers. Example: 7.8, 45.56, 5 point 6 Best Practice: Use a float and not number when decimal points are always discussed. Use a float and not money when you have to extract a currency amount, when there is a possibility that the unit may not be discussed in the turn.
Free Form	Entity which returns a string which is either a suffix or a prefix to a configured key word. Best Practice: We need to leverage AI intent model instead of Free Form Entity type.
Location	Entity which is a geopolitical entity, i.e. countries, states, cities. Best Practice: There may be many instances of the location, and it is the keyphrases that limit the false positives of a possible location hit. Hence, we need to be careful about the keyphrases configured.
Intent	Entity which contains intent of the call. Example: Claim status, Hotel booking, Change an address, etc.,
Money	Entity which contains monetary values, including unit. This entity type currently supports US Dollars, Indian Rupees, Korean Von, Australian Dollar, British Pounds, Euros and Japanese Yen. Best Practice: If the currency using is not spoken in the context then currency entity is not detected. Summaries should leverage rules where money and number entities are combined for such cases.
Number	Entity which contains numerals that do not fall under any other entity type.
Ordinal	Entity which contains a number that indicates the position or order of something in relation to other numbers.
Organization	Entity which contains the name(s) of companies, agencies, or institutions.
Percentage	Entity which contains any values in percentages. Example: 50%, 60 percentage, etc.,
Person	Entity which contains name of the person.
Product	Entity which contains product names.
Quantity	Entity which contains indefinite amount or number.
Relationship	Entity which contains relationship related key words. Example: Father, Spouse, Friend, Customer, Manager, etc.,
Rule	A rule is a set of conditions applied on NLP entities, AI entities, Complex entities or other Rule entities. When the condition is satisfied, then the configured output is calculated as the output of the rule entity. Best practice: Before each rule is created, we need to think twice if it is absolutely required. Accuracy improvement and tuning is much easier and cleaner when entities defined independently are used in summaries.
String	Entity which returns a string with a match to configured keywords. Best Practice: Fuzzy match will not work fully Avoid using a lot of string entities. It leads to performance degradation of Summary generation. It is recommended to use AI entities for these use cases.
Time	Entity that contains a time reading. Best Practice: It is recommended to customers about training agents in such a way that dates and times are always spoken by agents in a prescribed structure for good accuracy in model detection.
Website	Entity that contains a website or web address. Example: google.com, wikipedia.org, www.google.com, www.facebook.com
Confirmation	Entity that contains speaker affirmation based on the key words like yes, yeah, okay, no, I don’t want to, etc., Interesting Fact: This is different from Boolean entity. This purely captures what was confirmed (Yes / No) by the other speaker.

Redaction Configuration

Administrator has an option to upload redaction configuration file. It prevents Agents from seeing PCI, PII data and customer specific sensitive information in the transcription.

Follow the below steps to upload the redaction configuration file:

Click Add New Config to add a new redaction configuration file. A new row is added in the Redaction Configuration section.
From the ISO Language Code drop-down list, select the language code based on the redaction configuration file to be uploaded.

Click Download Template () icon to download the redaction configuration file (.zip). Extract the .zip file into the desired path. You can see the below JSON files:

customer_config.json
pii_all_custom.json

Sample Redaction Configuration File:

customer_config.json

{  
"entity_list": [],  
"allow_list": []
}

pii_all_custom.json

{	
    "recognizers": {		
        "PASS_CODE": {			
            "value": {				
                "entity": "PASS_CODE",				
                "language": "en",				
                "context": [					
                    "member"				
                ],				
                "patterns": [					
                    {						
                        "name": "code-regex",						
                        "regex": "(\\b\\d{10}\\b)",						
                        "score": 0.5					
                    }				
                ]			
            }		
        }	
    },	
    "analyze_template": {		
        "name": "analyze_pci",		
        "data": {			
            "fields": [				
                {					
                    "name": "PASS_CODE"				
                }			
            ]		
        }	
    },	
    "anonymize_template": {		
        "name": "anonymize_pci",		
        "data": {			
            "fieldTypeTransformations": [				
                {					
                    "fields": [						
                        {							
                            "name": "PASS_CODE"						
                         }					
                    ],					
                    "transformation": {						
                        "replaceValue": {							
                            "newValue": "<PASS_CODE>"						
                        }					
                     }				
                 }			
            ]		
        }	
    }
}

In the customer_config.json file, the fields entity_list and allow_list are blank by default. It helps to redact all the default system entities. For example, Credit card number, CVV number, Card expiry date, etc.
In the pii_all_custom.json file, mention the customer specific entity values that need to be redacted.
Click Browse. The window to select file opens. Select the redaction configuration file (.zip). The file name is displayed in the corresponding row.
Click Upload () icon to upload the file.

In this section:

Uniphore Customer Portal

System Setup

App Profile Settings

Ghost Call Configuration

Note

Filter Conditions

Note

Note

Configuration of filtering criteria

CTI Language Mapping

Add CTI Mapping

Entity Catalog Controller

Important

Note

Warning

Create Entity Catalog

Built-in Entity Types

Redaction Configuration

Search results