Named Entity Recognition Tool - Enough Data Points for Each Label

[hellyars]

I am trying to learn how to use the NER Tool in the Text Mining suite.

I am trying to figure out Train w/ New Entities. I keep generating the following error.  No tool example is provided.

I have a simple Text Input tool connected to the E anchor that contains 2 columns: Entities and Label (with Label containing only one value 'Program'). 

Error: Named Entity Recognition (6): Some labels don’t have enough data points to train an accurate model. Each of these labels has less than 20 data points: ['PROGRAM', ''].

[rizistt]

I ran into this issue as well. As a solution, I provided different variations (lower case, upper case, title case, and mixed case) and used them multiple times. It's not a solution but a way to make things work. I hope the Alteryx team will make it more usable in the near future.

[dhouse]

The "Train with new entities" feature of the Named Entity Recognition tool uses the data you provide to train the algorithm to recognize the context your entity generally appears in. There needs to be a minimum of 20 different examples of the context for each entity you would like to train, in order for the algorithm to learn.

[hellyars]

@dhouse 

Given the tool lacks any examples, that might be something worth mentioning on the tool's help page.

That said, I have more than >20 "entities" assigned to one "label" and it still generates the same 20 data point error.

[dhouse]

Could you share the entities that you are passing in?  Also, I have attached the NER one tool example (it will be available in the 22.3 release in the tool drop down).  

[hellyars]

@dhouse  Thank you.  Will do.  Dealing with a Windows - Alteryx icon issue after deleting an older version of Alteryx.