Entity extraction and multiple rows in an excel file (list rows presented in a table)

[Anonymous]

Hi all,

I've trained a custom Entity Extraction AI Builder model based on my data and it is working well.

I'm trying to scale up and use an excel with about 500 rows and in each row there is the text I wish to run the entity extraction model against.

The flow runs, but I'm still missing the cells (in update a row)  with the entity value and score. What should i change maybe in the below flow?

It also seems that i can't avoid the 2 loops below. The second is being created once i add dynamic variables to update the excel row.

Thanks a lot,

Itamar.

[CedrickB]

Hi Itamar,

I'm not 100% sure to get your question properly so let me rephrase.

You want to store, in a Dataverse table, the results extracting entities from text stored an Excel file.

Is this correct?

From your screenshot, you are using the Excel input in the "Update a row" action instead of the output of the entity extraction model. Is it intended?

To the second question, yes these 2 loops are created first to iterate over the Excel rows as input and then over the entities returned by the model. You can't avoid this.

[Anonymous]

Many thanks Cedrick!

I'm getting an excel file from my DWH with 2 columns (ID, Text) and wish to update this table with additional 2 more columns next to the 'Text' column. I do have a step in my flow that creates a table and add 2 columns which i can see in the file.

The issue is that none of those cells are being updated with the entity score and type. I'm not trying to update a Dataverse.

I've also tried to use the entity as in the below, but this also didn't work for me.

Thanks a lot,

Itamar.

[CedrickB]

It may be because no entity is extracted by the model. 

Did you take a look at the flow run history by clicking on the date on the last execution?

If so, did you click on the "Extract Entities from text" action to see the response of the model?  

If empty, can use the "quick test" in the model details page of the wizard to confirm that it works fine? 

[Anonymous]

I can confirm that the model is running OK when i test it and also when using the pre defined template of processing text from an email. Thanks a lot for those steps!

I think i also maybe found the error following your steps. Do you know maybe what does this error mean? 'No Rows Found With Id 8'

Thanks a lot,

Itamar.

[CedrickB]

It looks like the insertion in the Excel file table is failling.

Are you updating the proper Excel table? 

[Anonymous]

I'm also thinking that there might be an issue with the table update. It seems that it forgets the table name and thus not recognize it. I've now hard coded the table name and will update here what the result is.

Have a great day Cedrick,

Itamar.

[CedrickB · Accepted solution]

Perfect, don't hesitate to mark this thread as "Solution Accepted" so other members of the community know they can find a solution on their problem here.

[Anonymous]

Thanks again Cedrick!

I think i'm almost there :-). I was able to see now records in my excel table, but the flow has failed and i don't know what the error is. seems strange that it failed but also continued to run. 

Any idea maybe? The error that i see is : 'Extract_entities_from_text_with_one_of_your_custom_models' failed. Seems to me that the flow again is 'forgetting' the dynamic value of the row for some reason.

My table:

I also found this error now :

[CedrickB]

Is it possible that you are trying to get a row in your Excel by id?

The message says that the 'id' column is null or empty while it shouldn't.

[Anonymous · Accepted solution]

I think we can close this thread 🙂 Thanks a lot for your help and time!

The flow runs and it took me now 4 hours to run on 430 rows and get a result. It still fails in the end but seems to me as performance issue and maybe i should run on a small batch. The excel table is populated with the entity type and score 🙂

Thanks,

Itamar.

[Anonymous]

Hi,

Just to update here that i was able to fix the performance issue by changing the 'Retry Policy' in the settings and now the flow runs in 8 minutes instead of failing or running 1.5 hours.

Below are the configuration i've made in the 'Get Row' and in  'Update a Row' settings:

[mhurston]

Thanks for the tip on fixing the performance issue. I was able to reduce my run time by more than 50%.