When base spaCy needs a lift

Using spaCy for named entity recognition works well, but not in every situation, especially when it comes to person names. However, thanks to Hugging Face you can use Google BERT models as an ML engineer (not as a data scientist), to easily increase person NER accuracy.

DISCLAIMER: spaCy can incorporate techniques similar to what I’m about to describe, so don’t count spaCy out, this article is simply an alternative method of incorporate the technique.

A Few Words on BERT and Hugging Face

Hugging Face describes itself as a community where we can “Build, train and deploy state of the art models powered by the reference open source in natural language processing.”

It’s a place to build models or use models others build — this last bit is especially important.

Google BERT (Bidirectional Encoder Representations from Transformers) is, without getting distracted by how it works, a Google-created technique for NLP. Check out this piece by the BERT architects, which digs a little bit into how BERT works. The short version is that models built using Google BERT work well. Yes, I know, that’s a ridiculous oversimplification but this article is about using BERT, not creating BERT-based models.

#python-programming #python #nlp #named-entity-recognition #superior person name recognition with pre-built google bert