In FineBI, synonym configuration is an important way to improve the accuracy of the Q&A system.
When using the Q&A system, business personnel often use some unique internal jargon or non-standard expressions, which may be inconsistent with the standard terms in the system. Admins can use Synonym Configuration to set synonyms for table names, field names, dimension enumeration values, and specific terms or phrases, helping the Q&A system better understand colloquial queries from business users.
The following table lists common application scenario examples.
Enterprise jargon
[John Smith] = Mr. Smith/Sir Smith
[Budget] = Water Level
[Store Name] = Terminal/House Number
"Sir Smith's sales"
"What is the R&D water level?"
Custom word
[New York, Chicago, and California] = Key Cities
[Variance] = Data Fluctuation
[Employee Count, Employee Growth Rate Compared with the Same Period, Employee Tenure, and Employee Duty] = Personnel Situation
"Sales growth rate compared with the same period in key cities this year"
"How is the personnel situation this year?"
Specific calculation & specific business calculation
[Sales – Production Cost – Non-Production Cost] = Profit
[Debt Amount ÷ Asset Amount] = Asset-Liability Ratio
"What is the profit in August this year?"
System-matched My and My Team
[Username System Parameter] = My
[Department System Parameter] = My Team
"My sales ranking this year."
"How is the sales of my team this year?"
Configuration notice:
When the questions asked by business personnel are semantically and textually unrelated to the standard terms in the system, for example, the scenario exemplified in section "Application Scenario", synonyms need to be configured.
No synonym needs to be configured when questions can be fuzzily matched. The following table lists some examples.
Combined with large language models (LLMs), the system can fuzzily match field names.
If the field name is FineBI Sales, users can directly ask "BI performance".
FineChatBI has a built-in similarity matching algorithm. The content with the similarity threshold less than 1/3 can be matched.
Fuzzy enumeration value matching: When the enumeration value is Dove Chocolate, users can directly ask "Dove".
Fuzzy synonym matching: When [Jack's Group] = Southwest Sales Group Two is configured, users can directly ask "Southwest group two".
1. Choose Intelligent Q&A Configuration > Synonym Configuration to enter the synonym page. 2. Add the analysis subject (for which synonyms are to be configured) to the data list. Select the subject after adding it.
3. Configure synonyms for the data tables, custom words, and parameters of the selected subject.
Synonyms can be configured for table names, field names, and dimension enumeration values of data tables.
For example, select Product Sales Details and click Add to add synonyms to the table name, field name, and dimension enumeration value respectively.
Field name - Sales
"What is the performance this year?"
"What is the revenue this year?"
Performance/Revenue
Field name - Personnel
"How many staffs are there?"
"Annual trend in number of employees"
Staff/Employee
Dimension enumeration value - PC
"How much do computers cost this year?"
Computer
Dimension enumeration value - China Railway Group Limited
"Annual sales trend of China Railway"
China Railway
1. Click One-Click Configuration for one-click intelligent configuration by LLM.
2. After One-Click Configuration, fine-tune synonyms according to the language habits in the enterprise. For example, the Store Name is customarily called Terminal within the enterprise.
3. Click Save.
By the setting of Custom Word, synonyms can be set for specific terms or phrases to improve the accuracy and responsiveness of natural language processing.
Scenario One: Non-Standard Internal Language
Set Standard Word to Variance for Data Fluctuation. When users ask "Data fluctuation of monthly sales in 2015", the system will recognize this question as "Variance of monthly sales in 2015" and then perform calculation.
Scenario Two: Simple Indicator Calculation
You can configure simple arithmetic operations between indicators.
For example, if the standard word Sales Amount – Cost is configured for Profit. When users ask Profit, the system will calculate the profit by subtracting the cost from the sales amount.
Scenario Three: Content in a Specified Scope
When users ask situations about xx, the system will answer the relevant dimensions and indicators.
For example, if standard words Employee Count, Employee Growth Rate Compared with the Same Period, Employee Tenure, and Employee Function are configured for Personnel Situation, the system will answer the configured relevant dimensions and indicators when users ask "Employee situation".
Scenario Four: Setting Preference Dimensions
If multiple time fields exist in the data, custom words can be used to configure the correct time due to the different time preferences of different questions. For example:
Number of Resigned Employees prefers the resignation date. Standard words Resignation Record Count and Resignation Date can be configured.
Warehouse Situation prefers the inbound date. Standard words Inbound Count and Inbound Date can be configured.
Scenario Five: Configuring Specific Enterprise Knowledge
If specific colloquial instructions exist within the company, custom words can be used to set corresponding fields for specific words. For example:
For Over Budget, standard word Cost Exceeds 100,000 can be configured.
For Key Cities, standard words New York, Chicago, and California can be configured.
Users can directly ask personalized questions using My/Our/My team, such as "My performance ranking" and "Our team's sales". Fields need to be bound to system parameters.
For example, the Medical Representative field specifies the employee ID of each medical representative, and the username parameter fine_username also specifies the employee ID of each medical representative. In this case, the parameter fine_username can be bound to Medical Representative. After these two are bound, when medical representative A logs in to the system and asks "my sales ranking", the system can automatically identify the identity of the representative and provide corresponding sales ranking information.
$fine_username
Username
$fine_display_name
Name
$fine_role
Role
$fine_position
Department position
Enter the Preloading Configuration page and refresh data for the configured synonyms to officially take effect.
The system allows synonyms to be import and export through Excel files, helping users in batch management and editing.
Configured synonyms can be exported as Excel files, helping users review, modify and sort out synonyms in Excel.
Modify synonyms and re-import them into the system.
If many fields exist and the enterprise has a knowledge comparison table, a list of synonyms can be prepared as an Excel file and imported into the system for quick synonym configuration. (You need to download and use the corresponding Excel template.)
1. Download the import template.
2. Modify the template file to prepare synonyms in this Excel file.
Change the sheet name to the name of the data table in the subject. If the subject contains more than two data tables, additional sheets should be added.
Delete the sample data from each sheet and fill in the correct information. (If more than one synonym exists, separate them by commas).
3. Check the synonym effect in the post-modified Excel file, as shown in the following figure.
滑鼠選中內容,快速回饋問題
滑鼠選中存在疑惑的內容,即可快速回饋問題,我們將會跟進處理。
不再提示
10s後關閉
Submitted successfully
Network busy