NER Training data annotation for Spacy v3.0

This video explains how you can quickly and easily annotate the training data for a custom NER pipeline project for free. In this video, we are using open-source software tecoholic.gith... and a few custom python snippets to generate the training data for the NER model.
Github: github.com/dre...
Watch this video to Train the custom NER model: • Train Custom NER with ...
Host your NER model in Huggingface Hub: • Spacy models in Huggin...
==========================================================================
Watch other tutorials like this:
Host your Spacy Model in Huggingface: • Spacy models in Huggin...
Semantic Search using Elmo : • Semantic Search using ...
Topic Extraction using Embeddings: • Topic extraction using...

[@asiasowa]

Thank you for this video! Very useful and helpful stuff! Keep up the good work! :D

[@2000coque]

Good video, so much tanks. You helped me a lot.

[@LuckyPratama71]

is there any annotation tool for the preparing spacy ner data?

[@totoedgar7487]

Thanks you for the amazing tuto. Can we annotat severals data in the same tIME ?

[@deepakjohnreji]

Thank you, Yes you could.

[@shanmuganathanramalingam771]

Hi Brother How to annotate automatically for large text data by this i have to do it mannually

[@AI_4214]

Thanks for the video. How do I know if the model is overfitting or underfittin.

[@deepakjohnreji]

You can always validate against a good dataset to understand your model performance

[@AI_4214]

@@deepakjohnreji Thank you!!

[@louatiyossr945]

hello , i'm doing an invoice data extraction project , i'm wondering if i'm going to choose ml-based approach ,i will need a dataset of text like the one in the tuturiol , but for a project we need large dataset not only just few lignes ,right ! if yes , how to gather it please ?

[@deepakjohnreji]

Hello Louati, if it's an invoice data extraction project I would suggest not to train specific models unless you aren't getting great accuracy with models like LayoutLM
huggingface.co/docs/transformers/model_doc/layoutlm

[@louatiyossr945]

@@deepakjohnreji layoutML also needs a labled dataset , i'm just lost how to gather dataset and label it or should i just choose rule-based approach .I'll be very thankful if you answer me

[@deepakjohnreji]

@@louatiyossr945 If the invoices are pretty generic I would say better go with existing cloud solutions offered by Azure or AWS, if its custom and hard for these services then its better to train a model using any of the recent architectures.

[@caankitrmehta2281]

Hi, can i use my sample pdfs to annotate and train the model? Which is better option this or lable studio.
Thanks

[@caankitrmehta2281]

Looking for free version for data annotation. Kindly guide

[@deepakjohnreji]

@@caankitrmehta2281 Lable studio would be a little heavier, if you have to use this tool you can simply convert the pdf to a text file and upload it.

[@SR-cm2my]

Thank you very much for this video. Step 3. Don't you think it should be i[1]['entities'] = [(0, 0, entity_name)] instead?

[@deepakjohnreji]

Thanks, I will look into it, could you share your file with me to check.

[@radiccalesmoral8011]

Hi! Thank you so much for the content.
I have modified the values that json file. How could I display the annotations again to check if everything is correct?

[@deepakjohnreji]

Thanks, you could simply use print option to validate or save it to some variable or file.

[@ahmeddanjumaharuna6350]

Awesome, i do have a question though, while using the tecoholic, what if a single word, has multiple annotations category, how do i annotate it? tried to annotate a single word to different tags but wasnt working. Thank you

[@deepakjohnreji]

True, that's a limitation here actually. you may have to duplicate for multiple instances and try out, or you could create some additional format in the training file.

[@KetsuiDesu]

Am i the only one getting almost empty data from the NER tool? It does not look like the text in the video. It only contains the annotations but no text

[@deepakjohnreji]

Is your issue fixed?