Skip to content

Dataset Management

As a Data Expert, you have the ability to pin questions to the suggested board for that dataset. This means that users will see these questions as the initial pinned questions for a dataset. By default, any question that is Verified is put on the suggested board, but if there are specific questions that you would like to pin, you can mark them with the Suggested tag. To do this, click the three dots right below the response you wish to mark as Suggested. Then, simply click Mark as Suggested. Note that the response must be Verified before it can be marked as Suggested.

Dataset Panel

Admin access is necessary to use advanced features in the dataset panel. Normally, when inspecting a dataset, only the Knowledge panel is shown. However, admin privileges give access to the entire Question Library, all data Assets, all Conversations, and curating training data with the Training tool.

Knowledge Panel Overview

The Knowledge panel allows users to access all dimensions and metrics the model has utilized. Each dimension and metric has a state tag, which can be one of the following: Unset, Verified, Invalid, Deprecated, or Review.

  • Unset: Assigned to AI-generated dimensions/metrics not yet verified. Users cannot set this state with an existing dimension/metric. Unset dimensions/metrics can only be used in the conversation they were created for. To use them in other conversations, they must be marked as verified.
  • Verified: Indicates the dimension/metric is valid and reusable by the model. This can only be set by the user. When a user creates a new dimension/metric from the Dataset Page it will be set to Verified by default.
  • Invalid: Automatically assigned when the SQL is invalid. Users cannot manually set this state.
  • Deprecated: Users can deprecate dimensions/metrics they do not want the model to reference. If the model had already used a dimension/metric before deprecation, it will appear as a deprecated conversation.
  • Review: Indicates a dimension/metric that has not been finalized. This state is typically used for dimensions/metrics under consideration.

The user can only mark an Unset, Review, Verified, or Deprecated dimension/metric as either Verified or Deprecated. All other states cannot be changed by the user.

Question Library

The Question Library allows admins to manage and organize questions related to the dataset. These questions can be used to quickly retrieve insights and ensure that common queries are easily accessible. Admins can add, edit, or remove questions, and can also tag questions for better organization.

Assets

The Assets section provides detailed information about the dataset's contents, such as tables, views, and fields. Admins can explore these assets to understand the structure and schema of the dataset, which is crucial for accurate data analysis and querying.

Conversations

Conversations allow users to interact with the dataset through a chat interface. Admins can create, manage, and review these conversations to ensure they are providing accurate and useful information. This feature helps in maintaining a record of all user interactions and the responses generated by the system.

Training

The Training tool enables admins to curate training data for the dataset. This involves editing existing training conversations or creating new ones by asking questions. The tool forwards questions directly to the analytics system, bypassing the chat phase, to streamline the process of training the system on new data. This feature is intended for qualified data experts to help the Numbers Station model to better understand the input data.