Improve NER Person Name extraction

[mauuuuu5]

Hi everyone, I am using the following NER Tagger to extract names from a document, but the model is not extracting the information properly.

NER Tagger Model Training – knime

NER Names.knwf (2.1 MB)

I beg if someone can help me or give some advice to improve the NER process.

Thank you

[Artem]

Hello @mauuuuu5

I can see you are trying to train your own NER model for Spanish. I do not know what kind of text you are working with, so perhaps it makes sense to train your own model. But have you tried Spacy nodes (Redfield NLP Nodes – KNIME Community Hub)?

Here I also attach a workflow I have created to showcase the capability of these node for Portuguese. I think you can take it as a reference, just select one of the Spanish models in Spacy Model Selector node and provide your texts.

Redfield NLP Nodes: Spacy example – knime

[mauuuuu5]

Hi thank you for your help, I tried the workflow guess that I should adjust some options to get the desired results

since the model is not picking up persons names

I wonder which of the following models will the the best to extract person names?

I also uploaded a reset version in knime hub in case you can look at the workflow

Redfield NLP Nodes: Spacy example – mauuuuu5

Cheers

[Artem]

Hello @mauuuuu5

I took a look at your workflow. I believe there are 2 reasons why you could not extract the names. First, you did not apply the Spacy NER node in the branch where you do tag filtering. Second, you used the wrong set of tags for filtering — the POS tags, while NER tags is a different category. Here I attach the updated version of your workflow and the screenshot of the Tag Filter settings.

And here is my output from Bag of Words Creator:

Fixed workflow:

Redfield NLP Nodes: Spacy example – artem