Text mining - Named Entity Recognition, how do you re-use the model?

[aproctor14]

How do you re-use a "model" that comes out of the Named Entity Recognition tool (Intelligence Suite tool, using 2022.3)?

1. I have created a model:

- It has training entities (dates, and people names), and it is applied to an accounting system's "description" column (to extract dates from the field). The model is output to the green box. It takes around 20 minutes to run, on the revenue side. 

2. I would like to re-use that trained model on the costs side, without re-running the tool for 20minutes.

3. When trying to re-use the model, I receive the below error message. This is probably the wrong approach, but how can the model from this tool be re-used?

Thanks!

Step 1Step 2

[alexnajm]

This looks right, except it looks like Step 1 outputs to "Model2.yxdb" whereas Step 2 tries to read "Model.yxdb"?

[aproctor14]

Sorry - should have mentioned, that was just an intermediate step to avoid writing over the model.

The Model.yxdb is still valid and pulling in (the file input tool is finding it / opening it etc.)

[alexnajm · Accepted solution]

You need to use it with a Predict tool as indicated in the documentation! Named Entity Recognition | Alteryx Help

[NeilFisk]

The path provided above is broken.  Where do I find the help file now?

[alexnajm]

Just search on Google in my view - Named Entity Recognition (alteryx.com)

[NeilFisk]

I found where it was within the Alteryx site, but it's not clear how the Predict tools allows one to use the results of a previous data set and model to be applied to subsequent ones.  To me, it looks like one has to re-run the model with the new data against the new entities already defined in the NER tool.

[alexnajm]

I am not sure I understand the question - the model output from the model before becomes the input to your other workflow with the new data.

[NeilFisk]

So you still have to run the NER to generate the labels each time.  I'm really not clear on how this becomes the input for new data.  This is my first entrance into using natural language processing.

[alexnajm]

If you have new labels that need to be in the model, then yes you do. Otherwise you can just use the model output with your new data and the Predict tool together

If you have more detail and a workflow maybe I can help further, but otherwise I am just spit balling!