Skip to main content

June 2025

June brings computed fields, native Word and email parsing, and self-service sign-up.

Highlights​

Computed fields​

A field in a validation template can now take its value from other fields instead of from the document. Use it for a total that should equal the sum of its lines, a reference built from two other values, or a date offset from another date. A computed value can be translated and checked against a lookup list like an extracted value, using the fields it was computed from as context.

Self-service sign-up​

New users can sign up through a guided wizard, and organisations that need manual onboarding can send a request form. You can also sign in with a link sent to your email address. Creating a new organisation now takes you straight to it.

Week of 2 June​

Prediction & processing​

  • A model per extraction step: each new extraction step gets its own AI model.
  • Classification labels: manage classification labels through the API or in a new table.
  • Grouping on pipeline steps: pipeline steps have a grouping field.

Import & document processing​

  • Word (.docx) parsing: .docx files are parsed directly and are now an allowed file type.
  • Merge all attachments: a project setting that combines all attachments on an incoming email into one document, instead of one task per attachment.

Interface​

  • Pipeline preview: a visual overview of what a project's pipeline does.
  • Length warning: reviewers are warned when an annotation is longer than its field allows.

Week of 9 June​

AI & data​

  • Segmentation sets: segmentation training data can be assembled into sets.

Validation​

  • Template editing: edit validation templates in the interface.
  • Data types: each field type has its own parsing, validation and export behaviour.

Prediction & processing​

  • Model assignment: assign models to pipeline steps in the interface.

Week of 16 June​

Validation​

  • Computed fields: described above.
  • Context lookup: a lookup can use the values of surrounding fields as context, which helps when the value alone is ambiguous.
  • Grouping type on groups: a template group shows its grouping type.

Fixes​

  • Single-character lookup values are handled correctly.
  • Translation no longer runs again when only the letter case changes.
  • Upload and email names are set correctly.
  • Dropdown menus no longer overflow their container.

Week of 23 June​

Prediction & processing​

  • Most confident prediction wins: when several predictions compete for the same field, the most confident one is always chosen.

Import & document processing​

  • Email (.eml) parsing: .eml files are parsed directly, including their attachments.
  • Long documents: extraction no longer has a maximum document length.

Administration & access​

Fixes​

  • The zoom level is correct on all documents.
  • Contents of unpacked archives are filtered correctly.

Week of 30 June​

AI & data​

  • Classification datasets: build classification training data from processed tasks.
  • Chunking for retrieval: the retrieval step splits source material into chunks, improving extraction on long documents.

Validation​

  • Signatures and checkboxes: validation screens have their own controls for signatures and checkboxes.

Notifications​

  • Extraction complete event: get notified when extraction finishes, next to the existing task status events.
  • Email templates: notification emails now use templates.

Fixes​

  • Microsoft sign-in works again for all tenants.
  • Email bodies are no longer cut off when displayed.
  • Empty notification subscriptions no longer cause errors.