Data Characteristics

Additional data characteristics are important when analyzing interaction logs or interpreting data from the reporting dashboards or determining the billing amount:

• Environments

• User types

• Session Mode

• Contexts

• Languages/Cultures

Environments

Conversational AI Cloud is set up with 3 environments:

1. Development (where one can edit the project’s content and interactions via the CMS)

2. Staging (for testing and approval)

3. Production (for usage by the end-users)

Interactions on the production and staging environments are both logged, and by default logged as session mode public and test (respectively). These environments may have different versions of the project database. Interactions on the development environment are not logged.

User Types

Users can have any kind of interaction with Conversational AI Cloud. Some of these interactions can even be initiated by Conversational AI Cloud itself/the website or app connected to Conversational AI Cloud, without the active engagement of the user.

We classify users as:

• Inactive: Users are 'Inactive' if their only interactions have been initiated by the website/app or Conversational AI Cloud itself. These generally comprise interactions of the types Event, DataRetrieval, and dialog start interactions as long as the dialog is only started and not continued

• Active: Any user who has interactions of other types, is an active user

Session Mode

Interactions can come from various places, and for various purposes – tests, monitoring, content developers checking content, visitors to the public website, etc. The session mode can be set in the request sent to Conversational AI Cloud.

We’ve defined several possible modes for the sessions these interactions take place in:

• Internal: Content Editors/Developers working and testing content–usually in the CMS but this can also be defined for use of people internal to a company testing on the production environment.

• Monitor: For the purpose of monitoring the response/performance of Conversational AI Cloud. This mode needs to be set explicitly.

• Test: This should be set explicitly and by default for interactions to the staging environment. Test questions to the public website can also be given this session mode

• Public: Regular interactions on the public site/app interface with Conversational AI Cloud. This is the default mode on the production environment.

• Undefined: Anything else.

Contexts in Data

Contexts can use parameters that are available from potentially 5 different sources:

1. Header information. available from the browser’s information, e.g. the ip-address, the devices, the set language, etc. This includes derived information such as geo-location.

2. Information determined by the website/app and passed along in the request for the interaction. This is related to header information but can be more extensive than just information contained in the browser headers.

3. Client back-office information, e.g. order info, CRM profile info, etc. which does require a customer ID in some form that has to be passed along into Conversational AI Cloud so we can connect with that back-office system. It does also require a tech project to set-up this back-office connection.

4. Client third-party profiling systems, e.g. any existing (website) add-on that profiles users (like DoubleClick) which holds information about the segment of the user or its interest. As these systems typically all have APIs, Conversational AI Cloud can connect to these data to retrieve information about user (given the client’s consent).

5. Public information such as weather or traffic information.

Any part of these options can be used to define Context Values that can be used in all Interactions. With many Dashboards, Context Values are available as filters to drill down into the data.

Languages/Cultures

In Conversational AI Cloud, projects can have different languages but also different Cultures. Cultures are language variation s such as US English, distinguishing it from UK English.

A project with just English ignores the available cultures (which is fine). Cultures are treated like languages; all data are linked to that language and presented with that filter by default.