Knowledge Banks

How to use and configure the Knowledge Banks for your end-users.

The Knowledge Banks are the section in which your content is stored and filed. The end-user would never know they were accessing particular banks of content but for the admin, it is a way of separating content. In the future, this may be how content is permissioned to specific users, so we would recommend thinking of the banks as perhaps free vs paid-for content.

There is no limit on the amount of content you can upload to one bank, initially, we recommend keeping fewer banks, each with lots of rich content. The content here is what the agents will use to ‘learn’ from in order to deliver accurate and informative answers for your end users.

How to Configure the Knowledge Banks

Under the Manage tab of the left-hand navigation bar, go to the Knowledge section, here you will be able to configure the Banks for your agents. Select Add New Bank:

Then give the Bank a name and short description - this is only used for the administrator and is not end-user facing.

Once saved, a Content tab will appear enabling you to add your Content into the Bank. Here you can input your content items. Content items can be uploaded in the following formats: PDF, MP3 or MP4, or MOV. You can upload or drag & drop your files into the bank, but you cannot upload more than 10 at a time.

When uploading you will see the file processing on the right-hand side.

Once your Bank files have been processed they will sit in an unpublished state on the left hand side, now you have the opportunity to manually review the file before publishing. Click on the three dots on the right-hand corner of the file and select Edit.

Under the Content Item Settings tab, you can amend the content Display Name, this is what the end-user will see when searching for content or looking at segments. You can amend the Description for that piece of content and add a thumbnail image. We recommend including thumbnails that represent your brand, as well as that piece of content, think about the topic, the speakers or the event and how you can represent this in the thumbnail. You also have the option to include the Display date, this is how content will be ordered in the front-end for users i.e. most recent content.

You have the option to exclude the content from the Discovery and Segments panel. Enable this if the content should be available to the agent for the formulation of answers, but should not be directly available to users via the Discovery and Segments panel.

Next Labels can be added to organize the content for your end-users (in the Content Discovery interface). Labels need to have been set up already under Knowledge on the left-hand navigation, and selecting the labels tab (reference this article for details on how to set up labels).

Metadata

The content metadata is used by the language model to understand the file's context (this is not viewable to end-users). The Known As and Summary fields are pre-populated however can be edited. Known As can be changed to explain the type of media file, e.g. Video or PDF for instance.

If the file is a PDF a Transcription tab will be seen next to the Content Item Settings tab showing the text directly extracted from the content you have uploaded.

Configuring Segments

If the file is an MP3, MP4 or Mov you will see a Segments tab next to the Content Items Settings tab. In the Segments tab, you will see the content broken down into chapters, or segments, these are chunks of content that each relates to one topic. When a new segment is created, it means the topic has changed inside of the content.

For each segment, you can choose a speaker, or add one by selecting the Speaker - unknown boxes.

Select the box and type the person's name who spoke in that segment, when you have added that person once it will automatically populate them for any other segments where they speak. They will also be added as a speaker in the dropdown list.

This is important because the end-user can start engaging directly with specific speakers in the agent chat i.e. ‘What does Niamh think about using Accounts?’ and they will get an accurate answer.

If you select the pencil button for each segment you can amend the transcript or update the Segment heading and summary. For example, if it has spelled someone's name wrong you can go and amend it.

If you want to double-check what that segment is referring to, in your content you can play the video segment directly.

If you find the transcription consistently gets a word incorrect, you can use the search and replace feature to update this throughout the whole piece of content.

Once you are happy with the transcription, segments, and speakers, click back to your list of content inside the bank. You can now start publishing these pieces of content to the Knowledge workflow for end-users to instantly engage with

Publishing Content

On each piece of content select the three dots in the right-hand corner and click Publish File.

Once Published you will see the green tick appear in the left-hand corner.

Within the selected knowledge bank, you will see additional tab headings: Prompts, Bias and Access Management.

Prompts

To configure Prompts under the Prompts tab, select Add Prompt Collection. For details on Prompts and how to configure them, refer to this article.

Bias

Under the Bias tab select Bias Assessment to see a summary of the Banks Data. This provides information on the key themes and concepts of the banks content.

The Weighted Concepts tab displays the concepts and the related segments, with the highest weighted concepts at the top and the lowest at the bottom.

Access Management

Access Management is managed through Agents. Users will have access to Knowledge banks through the agents they are entitled to use. Users will also have access to this content through the Content Discovery interface if enabled.

The type of bank configured is visible in your administration portal above the Knowledge bank name (e.g. Open access content bank).

Please note: The default Knowledge bank in your application will always be an 'open access content bank'.

The People tab will show you all of the speakers you have added in your segments.

Within the list of speakers, you can edit their profile by selecting the three dots by their name, followed by 'edit'.

You can add their Profile URL (LinkedIn, social, or other website URL for this person). When users are chatting with the agent people (speakers) will be related to the suggested content segments and will link to any associated Profile URLs.

Discover

The Discover tab will show you how the Content Discovery feature will look to end-users, so you can test it on the admin side.

Labels

The Labels tab enables the configuration of labels, these can be added to help organise your content for your end-users. To find out more about setting up labels, please read our article on Setting up Content Discovery carousels here.

Content bank vs synced bank

There are currently two types of knowledge banks - content banks and synced banks.

The difference is in the way files can reach the bank.

Content bank - files can be uploaded directly into the bank.

Synced bank - files are ingested and synced from an external source such as Gdrive, via an ingestion integration.

Files cannot be deleted from a synced bank, however, edits can be made to the metadata of the files such as display name, description, and summary. Any changes made to the metadata will be honored during the syncing process.

In order to find out more about integrations, please go to our integrations article.

Last updated