Creating a Knowledge Base Response
This section of the user guide is focused on understanding how to generate and create knowledge base responses. To help you navigate the article we have broken it down into the following sections:
- Question Response Overview
- Who can create/edit a question response?
- Creating or editing a question response is available to individuals with the following permissions:
- Content Generator
- Hyperlinking within a response
- Variables
Question Response Overview
The response field is what will be displayed to the virtual assistant user when asked the question or another variation of the question. There are many ways a student may ask a question that will result in the response you are providing. Refrain from using yes/no responses or responses that don't provide context to the question being answered.
Additionally, the way a question is asked may be applicable to multiple offices. Review the question to see all of the ways it may be applicable on your campus, and adjust your response to accommodate all options. You may want to consider if the response should have all responses within the one response field, or if a clarifying question should be used.
For information on clarifying questions, review the Creating Clarifying Questions article.
To help you generate a response we provide you with a content generator to assist you in crafting a response and variables to help you maintain your content efficiently.
Who can create/edit a question response?
Creating or editing a question response is available to individuals with the following permissions:
For more information on user permissions, review the User Roles & Permissions article.
Content Generator
What is the Content Generator?
The Content Generator is a cutting-edge tool powered by ChatGPT API to help institutions efficiently create, edit, and update their knowledge base content responses for their virtual assistant.
The tool also allows for the use of web knowledge sources to search a predefined set of web pages belonging to the institution, providing a wider range of information and sources for generating responses.
How to use the Content Generator?
Under the Virtual Assistant > Knowledge base section, select a question from General Library, Custom Questions, Question Templates, or Awaiting Review, or select Create Question to access the content editing screen.
Under the Response section, select the Speed up your content creation with our Content Generator powered by GPT toggle.
In the Content Generator, there are four parameter options:
- Verbosity: this setting affects the length of the generated response. Moving the slider left generates a shorter and more concise response. Moving the slider to the right generates a longer, more wordy response.
- Creativity: this setting controls the confidence level of responses generated. The lower setting on the left yields more predictable responses and the right yields more creative responses that may not be as accurate.
- Web Knowledge Source: When web knowledge sources are configured and the setting is on, the tool searches the question and matches the data found on the web pages and then utilizes the data scraped from those web pages to generate the response.
This setting defaults to toggled on unless web knowledge sources are not configured in the virtual assistant. If web knowledge sources are not configured, then a warning message displays.Note: The content generator leverages web knowledge source content to generate a response. The web knowledge source data used by the Content Generator is the same data used for suggested links. Therefore, if a domain is removed, then it will no longer be used for the link suggestion function or the Content Generator.
- Existing Response: A toggle to indicate if the current answer should be utilized as a starting point or not.
Parameter settings are preserved per user session.
At the bottom of the Parameters section, select Generate Response to generate a response based on the selected parameters. Selecting Generate Response again will generate a different response. Adjust the parameters as necessary to generate the desired response.
Although the tool prioritizes content harvested from web knowledge source crawls, GPT may generate responses from a large non-institution specific data set.
Therefore, when the tool is used, it is best practice to proofread and iterate on the response to ensure the response accurately reflects the school's practices and policies.
Note: Be sure to review any links to verify their accuracy. Once verified, remove the following tag from the response: - {Link might not be working, confirm or replace} or you may need to update the text to reflect the link being used.
Additionally, review the response for any information that could be replaced by a variable. This will help you maintain your content more efficiently.
For information on using variables within a response, review the Variables section of the Creating a Knowledge Base article.
The user can manually make changes to the response in the text editor. In the Response text editor, select Undo to revert to a previous version of the response within the same session.
For best practices when using the content generator feature, review the Content Generator Best Practices article.
Hyperlinking within a response
URL
- If you paste the URL directly into the Response field, the URL will not be hyperlinked (or "clickable") within the virtual assistant. To hyperlink a URL, complete the following steps:
- Within the Response field, select the insert URL icon on the text editor.
- Insert the URL into the Link URL field and select the checkmark.
- The URL will now be underlined and in blue text.
- If you paste the URL directly into the Response field, the URL will not be hyperlinked (or "clickable") within the virtual assistant. To hyperlink a URL, complete the following steps:
Email
- To hyperlink an email, enter “mailto:” before the email address (no space in between “mailto:” and the email address).
- The email link will now appear in blue text and is now reactive (clickable).
Phone number
- To hyperlink a phone number (to allow mobile users to make a direct call by clicking on that number), use “tel:” before a phone number (no space in between “tel:” and the phone number).
- The phone number will now appear in blue text and is now reactive (clickable) for mobile users.
Variables
How Variables Are Used
Variables are used in the responses to general library questions to ensure that whenever possible, your custom terms and URLs flow throughout the general library knowledge base, thus reducing the need for further customization.
When creating responses from scratch or using the content generator, you may want to update terms or web pages with variables to make them easier to maintain.
On the Edit Question page, variables are identified by an exclamation point as the first character of the text.
Using a Link Variable in a Response
Link variables have to be used as a hyperlink within a response. A link variable can be used to hyperlink to text or can be combined with a text variable.
Variables can be stacked to show the text from a text variable and then hyperlinked to a link variable to make it reactive (clickable). Example: When you add a link variable to a response, enter the link variable (in this case, “!grant-link”) and then use the hyperlink tool to link the text to the link variable.
Example: Display the text from a text variable (in this case, !academic-calendar) but hyperlink the text to the link variable (!academic-calendar-link) that contains the URL leading to the academic calendar.
Further information about variable types can be reviewed in the Virtual Assistant Variables and Creating Custom Variables articles.
Office Contact Variables
The only variables that are not visible and editable in the admin portal are the office contact variables.
Please contact your Client Success Manager or open an Ocelot Support Ticket to adjust any of your office contact variable values.
There are two types of office contact variables:
Office-clarifiers
- Office-clarifier variables will drive the virtual assistant to ask which office the user would like to engage in before displaying the chosen office’s information.
- Office-clarifier variables have the "fancy-bracket" prefix !{office}.
- The variables below are the office-clarifier variables:
- !{office}-office - Displays the office name based on which office the user chooses.
- !{office}office-hours - Displays the office’s hours of operation based on which office the user chooses.
- !{office}contact-number - Displays the office’s phone number based on which office the user chooses.
- !{office}contact-location - Displays the office’s address based on which office the user chooses.
- !{office}contact-link - Displays the office's email based on which office the user chooses.
- In the general library, this variable is often "hidden" behind the word "email" as the email's hyperlink:
Alert: Since the functionality of these variables affects all offices using the virtual assistant, office variables should not be deleted when customizing your response.
If they are deleted during the customization process, Ocelot will not be able to make the question live and will work with you to provide a better response.
We recommend having an Administrative-level user lock the questions that contain these variables.
For more information on how to lock knowledge base responses, review the Content Locking article.
- On the General Library page, search "!{office}" to find the applicable questions that contain these variables:
Office-specific
- Office-specific variables will display the office’s contact information to the user.
- Office-specific variables have an acronym for each office as the prefix.
Example: "FA" is the Financial Aid prefix, which will display like: “!FA-OFFICE”
- In the general library, the contact-link variable is often "hidden" behind the word "email" as the email's hyperlink.
- If the office you are looking for is not listed, it is likely because there are no distinct variables for that specific office.
Details about these variables are explained in the Virtual Assistant Variables article.
Was this article helpful?
That’s Great!
Thank you for your feedback
Sorry! We couldn't be helpful
Thank you for your feedback
Feedback sent
We appreciate your effort and will try to fix the article