Joe Barrow

My name is Joe Barrow, and I'm an NLP PhD student at the University of Maryland working for Professor Philip Resnik and Professor Doug Oard.

Series Tracker

AllenNLP the Hard Way has 7 posts and is ongoing
Literature Review has 1 posts and is ongoing

AllenNLP the Hard Way

This series is my AllenNLP tutorial that goes from installation through building a state-of-the-art (or nearly) named entity recognizer. By the end we will have covered dataset readers, Hierarchical LSTMs and BERT, predictors, debugging, and experimental repeatability.

All the code for the tutorial can be found in the associated Github repository.

0

Part 0 - Setup

Covering Installation and Setup (including best practices)

1

Part 1 - Building a Dataset Reader

An introduction to Named Entity Recognition (NER) and the CoNLL'03 Dataset.

2

Part 2 - Building a Baseline Model

Building a sequence-tagging LSTM for Named Entity Recognition (NER).

3

Part 3 - Configuring Experiments

Combining the dataset readers and models to run our first experiment.

4

Part 4 - Tackling Your Own Experiments

Recommendations on how to tackle your own experiments.

5

Part 5 - Building a Predictor

Exploring AllenNLP Predictors for using your trained models.

6

Part 6 - Debugging

(In Progress) Commong debugging tips, tricks, and tools for AllenNLP and PyTorch.

Literature Review

This series contains summaries of interesting papers.

J

Jointly Learning Sentence Embeddings and Syntax with Unsupervised Tree-LSTMs

Summary of paper by Jean Maillard, Stephen Clark, and Dani Yogatama