Named entity extraction (NER)

The task in Named Entity Recognition (NER) is to find the entity-type of words. Entities can be locations, times or names.

Intros

• Introduction for named entities using Python at depends-on-the-definition.com
• A Review of Named Entity Recognition (NER) in towardsdatascience
• Named entity extraction Github - sample code for 3 different methods
• Papers with Code on Named Entity Recognition

Methods (from simple to advanced)

• Conditional random field (CRF)
• LSTM
• Bidirectional LSTM-CRF Models for Sequence Tagging paper on arxiv and article here
• LSTMs With Character Embeddings
• Residual LSTM network together with ELMo embeddings (paper here)
  • ELMo word representations are functions of the entire input sentence
• NER with BERT - highest score

Service offerings

• Azure Entity identification and entity linking
• Google Cloud Api Demo which also figures out phone # etc. You can also train custom models.

Client libraries

Good introduction O’Reilly Deep Learning for Natural Language Processing

• Libraries:
  • Open source - NLTK. Introductory exercise here in O’Reilly book here
  • spaCy Industrial Strength NLP
    • Article that shows custom NER on top of spaCy
  • CoreNLP / Stanford Python library
  • Gensim
  • Textblob
  • Flair A very simple framework for state-of-the-art NLP

Data sets

• Training data set entity annotated corpus
• Conll 2003

My own prototype

• Implementing spacy together with Flask / python, and deploying to Azure. Live website

Continuation 11/19/2023 with LLMs

• Spacy LLM