Active Learning Approach retraining of the model

Almost there!

So, I have 3 different types of documents flowing through this workflow once running. Is it possible to three different types of taxonomy in the same project? Or do I need to set up queues for each one and then a training project for each one? Each doc type is in the same modern DU project.

In the multi assign after the get transaction item activity I am seeing the error “SpecificContent” cannot be used like a method this is after copy and pasting the expressions in your tutorial.

Again any insights would be appreciated!

Thanks again for all your help Jon!

It is indeed possible to combine all 3 into a single performer, but that increases the challenge as you’d need to determine which document type the training data came from.

I’d suggest, as a start, keep them separate and use 3 different queues and 3 different performers unless you understand my tutorial extremely well and can take on the challenge.

As for your SpecificContent error, I need more detail, both of the error and the expression you are using.

Somehow I did not get an email that you responded.

Anyways I went ahead and did 3 queues with 3 different projects to handle each queue.

Where I am stuck at now is the last two casts for the documentDOM and the humanValidatedData.

When I tried your suggested approach I was getting the specific content error mentioned previously. I then changed from
queueItem.SpecificContent(“Human Validated Data”).ToString()) to
queueItem.SpecificContent[“Human Validated Data”]

That is at least getting me the value associated to that key. If I leave it with the ToString approach at the end I get a cant assign from type object to Extraction Result.

I then tried casting with (UiPath.DocumentProcessing.Contracts.Results.ExtractionResult)Newtonsoft.Json.JsonConvert.DeserializeObject(queueItem.SpecificContent[“Human Validated Data”].ToString())
and I get a runtime casting error.

That is where I am now. The last question I have is once we get these kinks worked out and I have the output from the “Machine Learning Extractor Trainer” activity I am not entirely clear how that gets back to the modern DU model.

Thanks for hanging in with me Jon!

You are building in a C# project, my code was written to demonstrate in VB.NET (I prefer C# myself but went with the most commonly used language).

You need to change the code I stated into C#, casting is different.

Jon, I marked your tutorial as the solution, however the only thing missing now is using the validated data to retrain the model. Far as I can tell their is no way to upload the validated data to the modern projects. There does not seem to be a link between AI Center and Modern DU Projects.

I’ll see if I can retrain mine soon using the documents I have been storing (cause at least we have them available to try to fine tune) and tag you in the tutorial with any relevant findings.

This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.