AI Functions / Text Anaytics

[MJEnnis]

I have created some really useful reports in PBI desktop using the text analytics. I missed the memo back in September that these features were going to be phased out. Fortunately, my reports kept refreshing until 14 December, much longer than the stated deadline. But then they abruptly stopped refreshing without warning. Since most of my reports in the service are based on one pbix file, everything is blocked. 

PANIC!!

This solution does not seem viable in my case, since I have 1000s of comments that have undergone language detection, keyphrase extraction, and sentiment analysis. 

I do not even know where to begin to fix this... Has anyone else had to fix this issue already? Any tips or tutorials anyone can share?

Please tell me I am not going to have to rebuild everything directly in the service...

[v-tejrama]

Hi @MJEnnis ,
Thanks for reaching out to the Microsoft fabric community forum.

I understand this situation was unexpected, and you are not the only one affected. The built-in Power BI text analytics features were fully retired in mid-December, following a brief grace period, which explains why your datasets continued to refresh temporarily before stopping. Unfortunately, there was no warning when the failure occurred, which has understandably caused frustration for many users who relied on these features.

On a positive note, you do not need to rebuild your reports or restart your work in Power BI. Your model, visuals, and measures remain intact; only the refresh step that depended on the retired text analytics capability has been impacted. Microsoft recommends using Azure AI Language services as a replacement, which offer similar language detection, sentiment analysis, and key phrase extraction. These services can be accessed directly from Power Query via the REST API or through Fabric, and will refresh as expected.

For cases where you have already processed large volumes of comments, the typical approach is to retain existing results and apply the new API to new or updated text moving forward. This helps manage costs and refresh times. If you need to reprocess all data, Azure AI can support that workload. As long as the output columns remain consistent, your reports and visuals will continue to function correctly.

While this change is not ideal, transitioning to Azure AI is the recommended and supported solution, allowing you to maintain your current reports. I hope this clarifies the issue and provides a path forward without needing to rebuild your work.

Best Regards,
Tejaswi.
Community Support

[MJEnnis]

Hi @v-tejrama,

Thank you for this reply! This is all clear to me, an amateur.

Yes, it was very frustrating that it happened suddenly without warning. Afterwards I saw that there had been a blog post back in August. But a warning directly in the app would have been very helpful! 

I have already proceeded with the first step you propose. To get the refresh to work again, I copied the exisiting text analytic data into an excel file and have queried it from SharePoint. I then deleted all credentials and query steps referring to the AI Functions. This way I have the historic data, but no text analytics for the future data. I needed to move fast because multiple reports and automated administration processes based on these reports were affected all at once! 😮 

Before deciding which way to proceed for the future data (API in desktop or connecting to AzureAI in Fabric) could you share some links to the appropriate documentation and tutorials? I knew where to find the information for the previous methods on MicrosoftLearn, but I do not even know where to begin now... 

Thanks again!

[v-tejrama]

Hi  @MJEnnis ,

Thanks for the detailed follow up and for explaining what you have already done. Given the pressure you were under, your approach of preserving the historical text analytics results and restoring refresh stability was absolutely the right call, and it is exactly what many customers ended up doing to unblock dependent reports and processes.

Since you are now at the decision point for handling future data, below are all the official Microsoft documents you need, collected in one place and grouped by approach so you can decide what fits best. I am listing them point wise as requested so they are easy to scan and refer back to.

Power BI and Power Query approach using Azure AI Language
This path is closest to how the old AI Functions worked and is usually the fastest to adopt if you are already comfortable with Power Query.

Azure AI Language service overview
https://learn.microsoft.com/azure/ai-services/language-service/overview

Language detection documentation
https://learn.microsoft.com/azure/ai-services/language-service/language-detection/overview

Sentiment analysis and opinion mining
https://learn.microsoft.com/azure/ai-services/language-service/sentiment-opinion-mining/overview

Key phrase extraction
https://learn.microsoft.com/azure/ai-services/language-service/key-phrase-extraction/overview

Calling REST APIs from Power Query using Web.Contents
https://learn.microsoft.com/power-query/connectors/web/web
https://learn.microsoft.com/powerquery-m/web-contents

Microsoft Fabric approach recommended for larger volumes and future scale
This option is better if you expect high comment volumes or want a more robust long term setup. Fabric notebooks handle batching and API limits more gracefully than Power Query.

Azure AI Language pricing
https://azure.microsoft.com/pricing/details/cognitive-services/language-service/

From a practical standpoint, many users start by implementing the API calls in Power Query so they can keep everything inside the existing PBIX and then move the logic to Fabric later if volumes or refresh times become an issue. Since you already separated historical and future data, you are in a very good position to do this incrementally without disrupting reports again.

Hopefully having all the documentation in one place makes the next steps feel much more manageable. If you run into questions while choosing between Desktop and Fabric, feel free to ask.

Thank you.

[v-tejrama]

Hi @MJEnnis ,

I wanted to check if you had the opportunity to review the information provided. Please feel free to contact us if you have any further questions.

Thank you.

[v-tejrama]

Hi @MJEnnis ,

I wanted to follow up and see if you had a chance to review the information shared. If you have any further questions or need additional assistance, feel free to reach out.

Thank you.

[Nabha-Ahmed]

Hi @MJEnnis 

I hope you are doing well!

You can replace the retired Power BI text analytics with calls to Azure AI services (Text Analytics REST APIs). These APIs provide language detection, sentiment, key phrases, etc., and you can:

Call them in Power Query

Or call them in Fabric (OneLake + Power BI connection)

This way you can still automate text analytics and refresh your dataset. Microsoft is recommending this as the replacement.

If this solution helped you, please give it a like 👍 and mark it as solution ✅ so others can find it easily.

[MJEnnis]

Thanks for your reply, @Nabha-Ahmed!

Could you share some links to the appropriate documentation and tutorials? I have already read on various posts that there is a way to do this in Fabric (see the links in my original post), but I do not know where to find explanations on MicrosoftLearn, etc. Thanks!

[Nabha-Ahmed]

You can replace the retired Power BI Text Analytics by calling Azure AI Language (Text Analytics) APIs from Power Query or Fabric. Below are the official Microsoft docs:
Azure AI Language – Overview
https://learn.microsoft.com/azure/ai-services/language-service/overview
Language Detection
https://learn.microsoft.com/azure/ai-services/language-service/language-detection/overview
Sentiment Analysis
https://learn.microsoft.com/azure/ai-services/language-service/sentiment-opinion-mining/overview
Key Phrase Extraction
https://learn.microsoft.com/azure/ai-services/language-service/key-phrase-extraction/overview
Calling REST APIs from Power Query (Web.Contents)
https://learn.microsoft.com/power-query/connectors/web/web
https://learn.microsoft.com/powerquery-m/web-contents
Microsoft Fabric Notebooks (recommended for large volumes)
https://learn.microsoft.com/fabric/data-engineering/notebooks

[MJEnnis]

This is really horrible by the way... this was such a cool feature that allowed non experts like me the opportunity to apply machine learning methods to my models. From the looks of things, it may take me weeks to months to learn how to and then replicate what I have already done... I am very disatisfied with this development.