AI-For-Beginners/translations/en/lessons/5-NLP/19-NER

Named Entity Recognition

So far, we have primarily focused on one NLP task—classification. However, neural networks can also tackle other NLP tasks. One such task is Named Entity Recognition (NER), which involves identifying specific entities in text, such as locations, person names, dates, chemical formulas, and more.

Pre-lecture quiz

Example of Using NER

Imagine you want to create a natural language chatbot, similar to Amazon Alexa or Google Assistant. Intelligent chatbots work by understanding what the user wants, often through text classification of the input sentence. The result of this classification is called intent, which determines the action the chatbot should take.

Image by the author

However, users may include parameters within their phrases. For instance, when asking about the weather, they might specify a location or date. The chatbot needs to identify these entities and fill in the parameter slots before executing the action. This is where NER becomes essential.

✅ Another example is analyzing scientific medical papers. A key task is identifying specific medical terms, such as diseases and substances. While simple substring searches might work for a few diseases, more complex entities like chemical compounds and medication names require a more sophisticated approach.

NER as Token Classification

NER models are essentially token classification models. For each input token, the model determines whether it belongs to an entity and, if so, which entity class it belongs to.

Consider the following paper title:

Tricuspid valve regurgitation and lithium carbonate toxicity in a newborn infant.

The entities in this example are:

• Tricuspid valve regurgitation is a disease (DIS)
• Lithium carbonate is a chemical substance (CHEM)
• Toxicity is also a disease (DIS)

Notice that an entity can span multiple tokens. Additionally, we need to differentiate between consecutive entities. To handle this, we often use two classes for each entity: one for the first token of the entity (commonly prefixed with B- for beginning) and another for the continuation of the entity (I- for inner token). Tokens that do not belong to any entity are labeled as O for other. This tagging system is known as BIO tagging (or IOB). After tagging, the title looks like this:

Token	Tag
Tricuspid	B-DIS
valve	I-DIS
regurgitation	I-DIS
and	O
lithium	B-CHEM
carbonate	I-CHEM
toxicity	B-DIS
in	O
a	O
newborn	O
infant	O
.	O

To establish a one-to-one correspondence between tokens and classes, we can train a many-to-many neural network model, as shown in the following diagram:

Image from this blog post by Andrej Karpathy. NER token classification models correspond to the right-most network architecture in this diagram.

Training NER Models

Since NER models are essentially token classification models, we can use RNNs, which we are already familiar with, for this task. In this setup, each block of the recurrent network outputs the token ID. The following example notebook demonstrates how to train an LSTM for token classification.

✍️ Example Notebooks: NER

Continue your learning with the following notebook:

• NER with TensorFlow

Conclusion

A NER model is a token classification model, meaning it can be used to classify tokens. This is a common NLP task that helps identify specific entities in text, such as locations, names, dates, and more.

🚀 Challenge

Complete the assignment linked below to train a named entity recognition model for medical terms, and then test it on a different dataset.

Post-lecture quiz

Review & Self Study

Read the blog The Unreasonable Effectiveness of Recurrent Neural Networks and explore the Further Reading section in the article to deepen your understanding.

Assignment

In this lesson's assignment, you will train a medical entity recognition model. Start by training an LSTM model as described in this lesson, and then move on to using the BERT transformer model. Refer to the instructions for detailed guidance.

---