Account Parameter Settings

See also: Standard language for account parameters at the organization level

AI Settings

AI training on the selected account parameter occurs when an invoice is posted. It means that the combination Supplier + Words in line description + Selected account parameter value is stored in the training memory and is thus included in the AI suggestions created the next time an invoice is received or when the invoice is restarted. Training does not affect suggestions for already received invoices. 

Training and suggestions:

• AI training only occurs upon posting of the invoice
• Suggestions based on new training are shown only upon receipt of a new invoice.
• To see suggestions based on training that occurred after receipt, you can start Reanalyze Invoice.

Suggestions and probabilities

• At a low number of similar invoices 
  When Semine has received a low number of similar invoices with the same supplier and line descriptions, the probability is often 100% for AI suggestions on the invoice. This is a significant advantage as it speeds up the process. The AI probability will decrease as more invoices with greater variation arrive.
  
  
• No automatic updated suggestions on already received similar invoices
  If many similar invoices arrive simultaneously from the same supplier before any of them are posted, the probability for AI suggestions will be identical for all received invoices. This probability will, by default, not automatically update as the invoices are posted and trained. If you want updated suggestions on unposted invoices, you can simply restart the invoice. 

See also: Settings - Accounting

Threshold Value for Auto-Application

This is the minimum % probability required for a suggestion to be automatically selected when processing an invoice. If the threshold value is set to 30%, the suggestion with the highest % probability over 30% will be automatically selected.

If there are 2 or more suggestions with the same highest probability over the threshold value, the user will have to manually choose one of them. For example, if we have 2 suggestions with 35% probability, the user will have to manually choose one of them. Semine AI will then train on the selected suggestion upon posting and thus increase its probability for the next time so that it will be automatically selected.

Override Existing Values

Here, you determine whether previous suggestions from rules, AI, or text analysis for other account parameters should be overridden by suggestions from AI for this account parameter.

Dimensions in Invoice List

Here, you can choose whether dimension values for selected dimensions (e.g., "Department") should be displayed in a separate column in the invoice overview. 
Here, dimension values (e.g., "Department B12") that are in use on at least one invoice line will be displayed.

Read more about Dimensional Display in Invoice List

Text Analysis

Extract Dimensions from Invoice Header

Extracts and adds the found dimension to all lines on the invoice. Also known as dimensional extraction.

Header fields extracted from EHF invoice:

1. cbc:Notes
2. cbc:AccountingCost
3. cbc:BuyerReference
4. cac:InvoiceDocumentReference
5. cac:ContractDocumentReference
6. cac:OriginatorDocumentReference
7. cac:ReceiptDocumentReference
8. cac:DespatchDocumentReference
9. cac:AccountingSupplierParty > cac:Contact > cbc:Name

Extract Dimensions from Invoice Lines

Extracts all fields in PDF and EHF/UBL invoice lines before scanning and replaces any dimensions set by findings in the invoice header.

• UBL/EHF invoice 
  All text fields from all fields in <cac:InvoiceLine> are extracted and scanned.
• PDF invoice 
  All fields on the invoice line except for VAT%, net amount, gross amount are extracted and scanned.

Also see full list of dimension extraction fields

Header Match Overrides Line Match

If there are matches on both the invoice header and the invoice lines, the former will be used for all lines.

What Values Are Searched For

Example of dimension value:

Code: 1234

Description: Finance

Alias: Economics

Include Dimension Code in Analysis (Search)

Here, SEMINE will search for the dimension code 1234 in the extraction, but this applies only when the code is a separate word with spaces around it. The code 123456 in the invoice extraction will not get a match even though the code 1234 is part of the word.

Note: It is recommended to have at least 6 characters in dimension codes to avoid false matches in the header or invoice line. A short code such as 123 may also match hotel room numbers and similar. A longer code such as 123456 is long enough to avoid false matches in the invoice extraction.

Include Alias in Analysis

Here, Semine will search for the dimension alias Economics in the extraction.
Note: This also applies if the search term is part of a larger word, such as Economics department. This is done to increase the number of matches.

Include Description in Analysis

Here, Semine will search for the dimension description Finance in the extraction.
Note: This also applies if the search term is part of a larger word, such as Finance department. This is done to increase the number of matches.

Minimum Number of Characters for Extraction

All words above a certain minimum number of characters are included in the text analysis. All words with fewer characters are excluded. This can help speed up the search. An empty field means there is no minimum limit for the number of characters per word.

Maximum Number of Characters for Extraction

All words below a certain minimum number of characters are included in the text analysis. All words with more characters are excluded. This can help speed up the search. An empty field means there is no maximum limit for the number of characters per word.

Ignore the Following Words

It can be useful to enter common small words (comma-separated) that you do not want matches on. Example: department, project, proj. This can help if you experience errors.
Example: If the description of the dimension value is "Dept on the second floor," SEMINE will search for all words in the line description. 

Do Not Split Words with Hyphen

• Setting Active: If this setting is active, words with a hyphen in the extraction will be considered as one, continuous word. In the extraction, there are words like "Hyphen" that will be combined into the word "hyphen" before Semine searches for matches.
  
  
• Setting Inactive: If the setting is inactive, words from the extraction with a hyphen will be split into individual words before the search is run. Example: In the extraction, there are words like "Hyphen" that will be split into the individual words "Hyphen" and "streak" before Semine searches for matches.

Overwrite Existing Values

• Active: If there is already a value set by AI, rule, or account parameter relation, this value will be overwritten by matches in the text analysis.
• Inactive: Previously set values will be retained even if Semine has matches in the text analysis.
  
  

Do Not Add Suggestions if Some Already Exist

• Active: If there are already suggestions from rules and/or AI in the dropdown menu, suggestions from text analysis will not appear in the menu.
• Inactive: Suggestions from text analysis will be listed in the dropdown menu along with other suggestions from rules or AI.
  
  

If Unique Code Is Found, Apply It Automatically

If the text analysis finds matches for only one dimension code, it will be applied automatically.

• Example: The text analysis finds matches only for code 1234 in the extraction, and this code will be applied automatically.
• Example: The text analysis finds matches for both code 1234 and code 3456 in the extraction. The codes will appear as suggestions, but none of the codes will be applied automatically.

If a unique alias is found, apply it automatically

If the text analysis finds matches for only one alias in the extraction, the dimension code associated with that alias will be applied automatically.

Auto-apply the lowest dimension code

If the text analysis finds matches for multiple alternative dimension codes, the lowest code should be applied automatically.

Auto-apply the highest dimension code

If the text analysis finds matches for multiple alternative dimension codes, the highest code should be applied automatically.

Auto-apply if the description has an exact match against the search text

If the text analysis finds a complete match between the dimension description and parts of the search text, the dimension value will be applied automatically.

Example
Dimension description: Department in the Basement
Search text: "Purchase of plants for department in the basement".
Here, there is a 100% match between the dimension description and parts of the search text.

Priority

Priority When Auto-applying

A numerical value that determines the order of execution for rules, AI suggestions, and text analysis for individual account parameters. Invoice enrichment runs in a fixed order, but priority determines the order of account parameters within steps in the enrichment process.

Fixed order for the steps in enrichment:

1. Rules (if in use)
2. AI
3. Text Analysis
   
   1. First found of
      1. Code
      2. Alias
      3. Description
4. Rules (if in use)

Example of priority:

We have 3 account parameters: NS, Department, Project. By setting priority for each of them, we can determine the order in which they are executed within each step of the enrichment:

1. Account
2. Project
3. Department

This results in the following final order for rules, AI, and text analysis.

1. Rules (if in use)
   1. Account
   2. Project
   3. Department
2. AI
   1. Account
   2. Project
   3. Department
3. Text Analysis
   1. Account
   2. Project
   3. Department
4. Rules (if in use)
   1. Account
   2. Project
   3. Department

So, if AI on Department (3) is set to overwrite existing values, then codes set (indirectly via relations) by rules for Account (1) or Project (2) will be overwritten by automatically applied codes set by AI for Department (3).

It is recommended to insert unique numerical values for each account parameter, as in the example above, to ensure they are executed in a specific order.