Models

• Out of the box models Improvement:

  • Splitting and classification has been improved further for receipt, invoice, and credit_notes

  • There are considerable improvements in the latest party information, especially in receiver name

• Line items Improvement:

  • Our latest model now better recalls tables and thus gives back better predictions

API

• Stats/inbox_id/evaluated_versions Feature: We have introduced a new parameter, which allows the user to indicate whether they want stats results based on the different version names (predicted, reprocessed, or other).

• Swagger/apidoc Feature: Apidoc now has a wider set of endpoints available for use.

UI

• Stats:

  • ML Awareness Feature: Our newest stats graph, Automation Insights, is up and running! This graph shows, based on the uploaded files, what the automation blockers are. When a file is uploaded it is either fully automated, or flagged for review for 3 different reasons, so this graph gives an overview of the 4 possible outcomes:

    1. correct: the prediction is correct and thus does not block the file to go STP

    2. thresholds: the confidence level of the prediction is lower than the threshold set, thus blocking the file

    3. prediction: if the prediction is empty, thus blocking the file

    4. computed fields: the computed field did not pass the test, thus blocking the file

• Review:

  • Data entry Fixed: Our date fields (data type: date) now always send back a uniform format. If a user then annotates a date that shows as “27.01.2022”, it will automatically be transformed as 2022-01-27. This is the same format as the JSON format also receives and sends it back so it will be the universal date format through out our BE to FE.

  • Actions > Extract a Table Feature: Annotating an extra large table and you made a mistake? We just added a “back” feature which allows for human mistake!

• Studio:

  • Formats - Computed fields Feature:

    • Computed fields can now also refer to header fields. For example, a computed field can now compare the sum of all items in a line-item column with a value given as header field.

    • Computed fields can now also be placed within line items.

• Settings:

  • Thresholds Improvement: We have now defined the optimized default threshold levels for the basic fields. These will be automatically set whenever those fields are added in the format.