High Performance Natural Language Processing

The HPNLP.org group consists mainly of enthusiasts, researchers, and students working on Natural Language Processing (NLP), who are interested in High Performance NLP algorithms and applications. The goal is to utilize modern technologies, e.g. GPUs, TPUs, distributed computation, to implement advanced NLP algorithms, probabilistic, connectionist, symbolic, for new high performance hardware. We are interested in miniaturization, mobile computing, and deep linguistic analysis.

Currently the group consists of faculty and students at Indiana University, and for Fall 2017 we intend to have regular colloquium meetings in this domain.

Current Team

• Damir Cavar
• Lwin Moe
• Ken Steimel

Example Projects

Resources

Sequential Python

Profiling Python

• python -m cProfile (script and arguments)

  Feature Extraction

• Scikit-learn

  GPU Processing

  Parallel Processing

• Joblib

  Distributed Processing

• Dask Cross Validated Grid Search
• Dask Distributed Schedulers
• Spark Machine Learning