CDI Language data in handled in ways which have undesirable impact on patrons in Primo, such as:
• A default language is applied to all records in a CDI feed for use in Primo in the Language field display, search, and facet. This may be done when the CDI provider feed has no record-specific Language information available automatically and the provider has not otherwise advised that all the records are in a specific known language and confirmed that this Language can be applied to the records in CDI
• Records may be allowed to merge into a CDI logical record even when the Language data differs across participant records. All of the participant record Language values are then included in the Language facet bar ‘outlier’ languages stripped out, and then one Language value may be selected to display in the record according to library full text activations, even if it contradicts the Language facet value

It is understood that the CDI is a massive index and that strategies such as collection level normalisation rules are necessary in order to manage this scale of data.
However, it is feared that the experience of individual users is inappropriately de-prioritised and lost in these decisions focused on “acceptable” percentages of impact, which may seem tiny relative to the size of the index, but may be significant to actual people using our discovery layer.
The practical reality is that such handling of Language information at a broad scale in CDI causes a negative impact for people using Primo, as the information is clearly incorrect and features cannot be used as expected.
It may be argued that patrons are unlikely to encounter such issues on a regular basis, but this also ignores common knowledge of human nature that even one or two experiences of being presented with clearly false information is enough to build distrust in all future experiences.

For example, the Language facet cannot be used when a patron is specifically trying to target results in a certain language, when they know in advance or learn from initial results that their search term may return records for multiple languages with different meanings.
ISSUE
“Persona” has different meanings in Spanish and English and the patron wants the English variation.
STRATEGY
With a search by “personas school principals” the patrons deliberately selects to include only the Language facet value of English and exclude the Language facet value of Spanish.
BUT
Ex Libris has applied Language facet value of English broadly as a default in the CDI provider feed and/or has artificially selected a Language from participant records to display in the record regardless of the Language in the facet.
RESULT
The patron sees in their results many records stating English when they are clearly in Spanish in both the metadata and full text.
Examples (screenshots attached):
• cdicrossrefprimary105944reopvol30num1201925194
• cdiunpaywallprimary1014295idonlinev15i583330
• cdijstorprimary_40654076
OUTCOME
This patron may then submit a complaint to the library or simply decide that their library provides a poor service and stops using our discovery layer in favour of other options.

When opening a case to report such patron complaints with examples of unexpected incorrect information in Primo, it is advised that this behaviour is expected and that current handling was chosen over other options to mitigate the impact of insufficient provider CDI metadata in Primo.
This could be by including a Language information in the records of ‘No linguistic content’ or ‘Undetermined’ to ensure a facet count matched to a result count, not allowing records to merge when there are Language differences, and not presenting Language facet values which do not match Language display values.

Ex Libris considers it a “resolution” to such cases for defect reports to document this behaviour, shifting responsibility to their library customers to explain to complaining patrons.

This approach focusing on quantity over quality does not centre the individual user as a person, and disregards the harm to the professional reputation of libraries, degrading the trust our patrons place in our services and which we are responsible to provide for our institutions, to support teaching, learning, and research.

Now that such behaviours have been documented over time as “expected”, this issue moves from hope it is resolved as a defect, to a product enhancement request that Ex Libris change the current design decisions causing this problem.

This submission requests that Ex Libris reconsider and change these practices with the individual user in mind, committing to align Language data across display, search, and facets and to consistently present only accurate Language information in the Primo discovery layer.