Learning Sparse Low Rank Tensor Representations of Transitive Verbs

This is a recreation and modification of

Fried, D., Polajnar, T., & Clark, S. (2015). Low-rank tensors for verbs in compositional distributional semantics. In Proceedings of the 53nd Annual Meeting of the Association for Computational Linguistics (ACL 2015), Bejing, China.
Chicago

---

Files

shell script to quickly upload output logs and saved weights to Google Drive using Skicka

• skicka_upload.sh

scripts to run train_parallel.py on cores using SLURM job scheduler

• slurm_run.sbatch
• slurm_test.sbatch

python scripts to train verb tensors with and without OpenMPI parallelization, respectively

• train_parallel.py
• train_verbs.py

simple utility functions used in verb.py and training scripts

• utils.py

contains Verb class, includes implementations of SGD and ADADELTA to train verb tensors

• verb.py

---

Requires

• Numpy
• Pandas
• tqdm
• SimpleMPI (for train_parallel.py)

---

Full report will be available soon.