This assignment will provide you with hands$-$on experience in using state$-$of$-$the$-$art NLP techniques for entity and relation extraction. It will also allow you to explore the capabilities of pre$-$trained language models for these tasks.

We are going to explore using both spacy alone and spacy $+$ Hugging Face Transformers.$$ As you have read in your book, Hugging Face API is a great resource for working with BERT and other LLMs$.$give python code.

For this assignment, you will use either dataset $($movie reviews or tripadvisor $($pick one that is richer in named entities$)).$ You will build a model that can extract entities and relations from text data. You will first experiment with spacy, and capture the results.$$ Then you will work with a pre$-$trained language model and spacy together, to capture the results.$$ You will then compare the results.$$ You can do all of this in CoLab.

What To Do

If necessary, install the libraries, including Hugging Face Transformers and PyTorch or TensorFlow.

Download the spacy model

Download the pre$-$trained language model using Hugging Face Transformers.

Load the movie review or TripAdvisor dataset.$$ There needs to be entities and relations present for a good extraction.

Preprocess the data, including tokenization and encoding, to prepare it for training.

Perform the entity and relation extraction using spacy

Perform the entity and relation extraction using LLM$+$spacy

Evaluate the performance for both entity and relation extraction tasks.

If you feel lucky:$$ Let's take this one step further and also perform an automatic knowledge graph generation $($Note:$$ It may not be very good but you will get a sense of how this works$)$ and visualize the graph.