Preloading Configuration

Application Scenario

FineChatBI allows admins to preset common analysis subjects and labels for business users, and set permissions. When a business user enters the query page, the left navigation bar automatically displays the preset contents within the permission range. You can click a content to ask questions about the selected data range.

Function Description

Preloading Configuration has the following advantages:

• Quick query response: The system preloads the preset subject data, reducing users' wait time.

• Quick selection of query data: Admins can add labels to the preloaded subjects, helping business users quickly select data to ask questions, and improving selection efficiency.

• Advance on learning dimension enumeration values: If admins do not enable Dimension Learning for the dimension field Material Number, when users ask "What is the inventory of AP-02-01?", the system will not yield a result. Preloading Configuration allows  dimension values to be learned in advance.

Adding Preloaded Subjects

1. Choose System Management > Intelligent Q&A Configuration as the admin, and enter Preloading Configuration.

2. (Optional) Click the Edit button above to adjust the max quantity of preloaded subjects. (Too many preloaded subjects may deteriorate performance.)

3. Click Edit, and add subjects as the preset data range when users ask questions globally.

Editing Subject Information

Click the Edit button next to Construction-Sales Analysis and edit the subject information (labels and introduction) for this subject.

• Label: Select a label from the drop-down list. (Labels can be available only after being created in advance. For details, see Labels.)

• Introduction: Add description information to the analysis subject for other users to select and use. You can manually input the description or use Intelligent Paraphrasing to generate one.

Note:

Admins can use Intelligent Paraphrasing only after configuring LLMs.

After the configuration is completed, you can see labels and introductions on FineChatBI query page.

Construction-Sales Analysis is under Sales Scenario. Click this subject to view the introduction of the subject, allowing users to understand the subject's data in the data list.

Learning Dimension Enumeration Values

• Basic learning: learning the first 10,000 data records (selected by default)

The specific logic is to deduplicate the first 10,000 data records obtained, and then learn dimensions based on the deduplicated results. Dimension learning may not be comprehensive.

For example, if Basic Learning is selected for the Store ID dimension and the first 10,000 data records only cover some stores, the system cannot accurately answer questions asked by users related to stores that are not included in these 10,000 data records.

• Full-volume learning: learning all data (which, however, will increase the performance burden)

The specific operation is to deduplicate all data, and then select the first 10,000 dimensions after deduplication to learn.

For example, if Full-Volume Learning is selected for the Store ID dimension, the first 10,000 dimensions after all records specified by Store ID are deduplicated will be learned, ensuring more comprehensive and accurate answers to questions about various customer names.

Note:

If you select Full-Volume Learning, you need to set Max Number of Characters. Dimension enumeration values with characters exceeding the upper limit will be ignored.

• Disabling Dimension Learning

When no query scenario is available for some dimension values, for example, specific user comments, Dimension Learning can be disabled.

If Dimension Learning is disabled, is any way available to ask questions about these dimension values? For example, if Dimension Learning is disabled for the Store ID dimension, when users directly ask "What is the sales volume of D010101?", the system will be unable to return results. This occurs because the system has not learned the Store ID dimension and does not recognize the dimension value D010101. There are several ways to solve this:

• Add keywords such as is, equals to, and belongs to to be used in questions like "What is the sales volume for the store whose ID is D01010?".

• Use quotation marks in questions like "What is the sales volume of where store ID D010101"？

Enabling Fuzzy Match for Dimension Values

After enabling Dimension Learning, admins can also set whether to enable Fuzzy Match for these dimension fields. The following similar scenarios can be achieved:

• Enable Fuzzy Match for the Customer Name field. When you ask "What is the sales volume of FanRuan Software?", the system will automatically match "What the sales volume of FanRuan Software Co., Ltd is".

• Enable Fuzzy Match for the Province field. When you ask "What is the sales volume of California?", the system will automatically match the sales volume of California state.

Refreshing Dimension Values

Since the data in the database will change, the dimension values in the preloaded data need to be refreshed to ensure that the BI Q&A can obtain the latest dimension values.

• Auto Refresh: Set the frequency of automatic refresh. The system will regularly obtain the latest dimension values according to the set frequency.

• Refresh All: Click Refresh All to immediately refresh the preloaded data in all data lists and obtain the latest dimension values.

• Refreshing of specific subjects: Click Refresh Cache next to a specific subject to immediately refresh the data of the current subject and obtain the latest dimension values.

Saving Preloading Configuration

After the above preloading configuration is performed, click Save for the configuration to take effect.

After preloading configuration is completed, the use permissions on the preloaded data need to be configured for relevant users. Users can only ask questions about the preloaded data on which they have permissions.

For example, user echo want to ask questions about the preloaded data of Chain Operation Analysis.

• Method one: User group dimension configuration

Choose Permission Management > Intelligent Q&A as the admin, and assign the use permission on the preloaded subject Chain Operation Analysis for the specified user echo.

• Method two: Resource project dimension configuration

Switch to Resource Project and configure editing permissions with the subject as the permission carrier.