Q&A not working

[mikolaj_rudnik]

Hello team!

I am facing an issue with Q&A functionality. 

I am working on the report in an import mode, which I can't share here.

Previously, this report had around 40 tables (including field parameter tables), but it was requested to merge this model with another one in an import mode. After the merge, number of tables is around 96 tables.

At first, there seemed to be no issues with Q&A feature, it was tested and no regression appeared.

I added new measures and fields from new part of merged model to Q&A, it was tested again and it was returning correct results from both parts of the merged model. But after few days, testers noticed that Q&A is not returning queries from the new model. When checked, it turned out that Q&A stopped working in the desktop version. It was greyed out in the visual and "Q&A setup" option in Modeling ribbon also was not working.

In service, Q&A looked like it was working, but it could not fetch any result, because it could not understand any field or measure names (even though they were added in synonyms).

I did some investigation on this forum and there were similar issues caused by Linguistic schema. In my case, there was an error when I tried to export Linguistic schema.

Because I was not able to export and see YAML file with Linguistic schema, I deleted measures and tables causing the error and added them again. After that Q&A feature started to work again, so I stopped there and did not investigate further.

Unfortunatelly, next day the error appeared again, so I repeated steps that worked last time, and Q&A was working again. This time, I knew that this is not a permanent fix, so I tried to export Linguistic schema. It was successful but with a warning: "Maximum number of entities or relationships was reached. Linguistic schema was truncated."

I was sure, that if the issue will appear in the future, simple import of Linguistic schema (that was exported from working file) will fix the issue. But the next day, when the issue returned, I imported Linguistic schema file and it didn't work. There were no changes in the file, so pbix file used for export was the same as the pbix that I was imorting file into.

My next step, was to try to fix Linguistic schema YAML file, but changing or removing terms from the error message had no effect -Q&A was not working and error still returned the same message (even though the terms were not present in the Linguistic schema files). 

I also tired clearing the cache in desktop report file, but it had no effect.

When I need to push any change in report to service, I have to manually delete measures and tables and add them again (which as you can imagine, is not a sustainable option) and then upload pbix.

After I close working report on desktop and open it again immediately, it still works. But when I open the same file the next day, the same issue with Q&A and Linguistic schema is back (even if there was no changes after publishing the report).

At this point I am stuck, because I could not find any solution on any forum or other Internet resource. 

If somebody faced simillar issue or has any idea what may be the resolution of this issue, I will be very grateful!

[tuphan]

Hi Mikolaj,

Sorry to hear that. Have you opened a support ticket yet? If not, I'd recommend you to do so. We will help you with this.

For this issue, it is likely that your linguistic schema contains an empty term which is not allowed in PowerBI. As you can't export the linguistic schema, I'd recommend to use an external tool like Tabular Editor to remove empty terms from the linguistic schema. The linguistic schema is stored in the Translations/en-US/Linguistic Metadata/Content property of the tabular model. You should use any text editor to update the linguistic schema and copy it back.

To update the linguistic schema to remove empty terms, search for "". If you will find something like this:

{"":{"State":"Suggested","Source":{"Type":"External","Agent":"PowerBI.VisualColumnRename"},"Weight":0.9}}

pls remove it and replace this process until there is no empty term left. 

Make sure that you back up your pbix file before you try this.

[v-linyulu-msft]

Hi,@mikolaj_rudnik 

Regarding the issue you raised, my solution is as follows:

1.Regarding your reference to reaching the maximum number of entity or relationship warnings, our analysis suggests that your model may be beyond the scope of the Q&A function due to its size and complexity. Power BI automatically creates language patterns based on the data set structure, but its processing power is limited. As you have tried, editing the language pattern directly in search of improvement is not guaranteed to work if the problem is rooted in the complexity of the model itself.

Here is the relevant documentation:

Limitations of Power BI Q&A - Power BI | Microsoft Learn

Edit Q&A linguistic schema and add phrasings in Power BI Desktop - Power BI | Microsoft Learn

2.Second, you can optimize the size of your data model:

Move unnecessary tables around on demand and store data using the direct query storage pattern.

View measure, relationships, and performance of visual objects: Review metrics in models, relationships between tables, and performance of visual objects. Make sure they are designed to be efficient and optimized.

Use variables to improve performance and troubleshooting: Using variables in DAX expressions can improve performance and help troubleshoot problems. Variables can store the results of calculations to avoid double calculations.

3.Finally, you can also try using incremental refresh. An incremental refresh strategy for partitioning tables, where only the latest imported partition is refreshed and another DirectQuery partition can be used for live data, can significantly reduce the amount of data that must be refreshed. At the same time, this policy ensures that the latest changes to the data source are included in the query results.

Here is the relevant documentation:

Incremental refresh for semantic models and real-time data in Power BI - Power BI | Microsoft Learn

Configure incremental refresh and real-time data for Power BI semantic models - Power BI | Microsoft...

Best Regards,

Leroy Lu

If this post helps, then please consider Accept it as the solution to help the other members find it more quickly.

[rajendraongole1]

Hi @mikolaj_rudnik - Might be some reasons like increased complexity and size of the merged data model and also check on current data model.

it get impacts the number of tables and the complexity of the relationships can impact the performance and functionality of features like Q&A.

you optimize it, by removing the unused tables , clear cache and optimzed relationship

Check the below link for more information solved:

 Solved: Issue with Q&A visual - Microsoft Fabric Community

If none of the above steps resolve the issue, consider contacting Microsoft Power BI Support for assistance.

Did I answer your question? Mark my post as a solution! This will help others on the forum!
Appreciate your Kudos!!