The Invoice Header Locator locates values for four different fields. Each field is represented by a tab in the locator properties window. On these tabs, you can add keywords and assign a weight to each keyword. For example, when searching for an invoice number, the keywords invoice number would likely have a weight of 100% whereas number alone might have a weight of 50%. You can also assign negative weights to words like Customer number.

Instead of keywords, you can use dictionaries that are defined on the Dictionaries tab in the Project Settings window.

Assign a weight of between -100% and 100% to a keyword. If a keyword is unique and can never lead to a wrong result, assign 100%. If the keyword must be found in combination with other keywords or might indicate a wrong result, use a lower weight. If different keywords are related to the same alternative, their added weights are used as the confidence level of the alternative. However, the confidence level never exceeds 100%.

Assign a negative value to use keywords as stop words. For example, give Order Number a weight of -100% if you are only looking for the invoice number.

Keyword tabs for the Invoice Header Locator are identical. The following example displays the Invoice Number Keywords tab:

Keywords can be single words or entire phrases. Both are searched with fuzzy logic and are case-insensitive, so you do not need to add keywords for all possible variations or errors.

Assign a keyword with a weight that signifies how important or how strong the keyword is. For example, define invoice number with a weight of 100% and “number” with 50%. Using these weights, if the Invoice Header Locator finds invoice number next to a number, it has a high degree of certainty that this is the invoice number.

However, if the phrase invoice number is not found because number: ABC1234 is used instead, ABC1234 could still be the invoice number. Even so, “number” might indicate the customer ID, so the locator cannot be confident that it found the invoice number. This is expressed by assigning a lower confidence level to this word.

Instead of entering a large number of individual keywords, use dictionaries. For example, you might have two dictionaries for invoice numbers, one with strong keywords like invoice number and one with weaker keywords like number. You can assign the weights to the entire dictionary, which lets you change the weights quickly.