Skip Navigation
Expand
Description of SmartAssistant
Answer ID 322   |   Last Review Date 07/22/2019

What is SmartAssistant and how does it help our customer base?

Environment:

SmartAssistant Suggested Solutions
Oracle B2C Service

Resolution:

When your customers can’t find an answer to their question in your knowledge base, they will submit a question to your support staff.
 
SmartAssistant is the term for an intelligent feature set that presents a set of answers that are relevant to the question the end-user is about to ask, before they are finished submitting the question. SmartAssistant is defined within incident business rules and a widget. This feature is intended to reduce the number of submitted questions to your knowledge base by providing end-users with the information they need prior to completing the submission process. 
 
The SmartAssistant "suggested answers" feature uses natural language processing to display a list of answers from your knowledge base. These answers are likely to contain the information your customers are seeking, as they are based on the question that is being typed. 
 
SmartAssistant can also display standard text or a single, specific answer.  SmartAssistant can  be used to redirect the end-user to a specific page when attempting to submit a question. SmartAssistant can even be configured to request that the end-user provide additional information before allowing them to continue with the question. 
 
If SmartAssistant displays information that addresses the customer’s concern, a submitted incident may avoided.
 
Customers also see SmartAssistant suggested answers and community discussions when they submit a question to the community. However, please note that incident rules do not apply to community content.

Related Answers:

There are two main technologies that underlie SmartAssistant. The first is a "Swarm Intelligence-styled" approach to data mining -- otherwise known as Related Answers. Related Answers are the links that sometimes show up at the bottom of an answer that you are reading on the end-user interface.

This feature of SmartAssistant continually monitors end-user sessions, gathering information about which answers your customers are finding most helpful, and tracking the path by which those customers are reaching those answers. When you click a link to view a Related Answer, you are given access to information that other users frequently viewed as well; thus, learning from the experiences of other customers who have had issues similar to yours.

Suggested Solutions:

A second technology utilized is Natural Language Processing, which is performed during searching and when submitting an Ask a Question request. The results of the Natural Language Processing are then compared against the published knowledge base to find the best matches.

This technology is used for SmartAssistant suggested solutions, both when configuring workflow rules to suggest answers in a preliminary response, or when staff members click the Suggestions... button when editing an incident.

Essentially, the subject and thread of an incident are language-parsed and fed through the searching algorithm of Oracle B2C Service. This content is compared to clusters of answers to determine which cluster the incident content best matches. A set of the closest matching answers from the answer cluster are returned from your published answers.

Reduce Solved Count of Unused Answers:

In addition, the agedatabase utility provides the ability to manage the size of the database by pruning old SmartAssistant information from the clicktrack (session tracking) and links tables. Agedatabase also implements a decay function on the score of both answers and links so that unaccessed information receives a lower ranking over time. As a result, answers that have not been accessed in a long time have their score reduced and are less likely to be listed as a related answer.


For more information, refer to the following resources:

Popular Answers About SmartAssistant

Using SmartAssistant on Submit a Question pages

Adding Content to Responses