Keeping the data clean.

Moliri saves every value, even when something is still missing. The Check area collects the open points: missing required details, similar spellings, references that lead nowhere and objects that may have been catalogued twice. Moliri finds and suggests, a person decides.

Your institution sets the rules and required fields. We set them up together with you.

Film: Checks and data upkeep
Film: required fields, data upkeep, duplicates and spellings in the “Check” tab

Required fields and rules in view

Check counts the objects with missing required fields, incorrectly filled fields and duplicate values, and one click shows them in the gallery. Above an object you see how many fields do not meet their condition, including those on other tabs of the input mask, and a button takes you to the field.

The server keeps the list of open points and updates it after every save. The numbers are there as soon as you open the area, and as a gallery filter, “Conditions not met” shows every object concerned, even in large collections.

The Excel import checks the conditions as well and records in a report which row breaks which rule. The data is written all the same.

Field conditions

  • Required fieldsObjects that lack a required detail
  • IncorrectValues that break one of your institution's rules
  • DuplicateValues that should be unique and are not
  • ImportReport per field, rule and row

One spelling instead of five

Moliri finds values that differ only in upper and lower case, umlauts or punctuation, as well as typing errors, swapped parts of names and abbreviations such as “Öl a. Lw.” or “Öl/Lwd.” (oil on canvas). It searches field values, authority data, locations and addresses.

If you wish, a language model looks for values that mean the same thing but are written differently, every night if you like: “Florenz” and “Firenze”, “Aquarell” and “Wasserfarbe”. Its groups appear in the same list, with a reason. If a group does not fit, you discard it, and it does not come back in this project.

You choose the right spelling, Moliri applies it to all objects concerned in one step, and that step can be undone as a whole. Duplicate authority records are merged, with both entries side by side and a trial run beforehand.

Similar values

  • FindsSpellings, typing errors, swapped parts of names, abbreviations
  • With AIAnother language, a synonym, with a reason
  • InField values, authority data, locations, addresses
  • StandardiseFor all objects in one step, can be undone as a whole

Finding gaps in the collection

Objects without a file, references to entries that no longer exist, and implausible values such as a date of creation in the future, an end before the start or a dimension of zero. Each list opens the objects concerned in the gallery.

Administrators also find authority records without a GND or Wikidata identifier there, locations without coordinates and files that are not used anywhere.

Data upkeep

  • No fileObjects that do not have an image yet
  • References to nowhereLinks to deleted entries
  • ImplausibleDate in the future, end before start, dimension zero

Finding objects catalogued twice

Moliri compares objects using the fields your institution switches on for this, and automatically using those that should be unique, such as the inventory number. It shows pairs that probably refer to the same object. The comparison is generous: “1921.1239” and “1921/1239” count as the same.

A pair that is not a duplicate you mark once as “Not a duplicate”. After that it no longer appears.

Suspected duplicates

  • ComparedFields switched on, unique ones automatically
  • GenerousPunctuation, upper and lower case, accents, leading zeros
  • Decision“Not a duplicate” is permanent

Cleaning up values on saving

Before a value is saved, it goes through a cleanup: spaces at the edges are removed, a decimal comma becomes a number. Your institution decides which steps apply per field, per project, per field type or for the whole instance, with a test that shows the value before and after.

For existing data, a trial run first counts how many values would change and how often each step took effect, with before and after examples. The cleanup itself is one step that can be undone as a whole. Old plain text in a JSON field can also be turned into entries this way.

Cleanup

  • ScopePer field, per project, per field type, for the instance
  • TestBefore and after on a value of your own
  • Existing dataTrial run with figures and examples, then one step that can be undone

An assistant suggests, a person decides

An AI assistant that you connect with its own access reads in Moliri with your permissions and can suggest values, such as a date taken from the description. It never writes into a field itself.

Each suggestion appears under Check and at the field, with its reason, its evidence and the model that made it. A person accepts, edits or discards it. An accepted suggestion appears in the history like any other change. The reason can be carried over as a source at the field.

AI suggestions

  • SuggestionCurrent value, suggestion, reason, evidence
  • DecisionAccept, edit or discard
  • TraceableMarked as an AI suggestion in the history

Check your own collection

Send us an extract of your data. In the demo we show which spellings, gaps and duplicates Moliri finds in it, and standardise one of them.