Skip to content
Our CL-AFF Shared Task Repo
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
CPU
cl_aff_utils
csv
scripts
.gitignore
README.md
TSNE_graph.png
TSNE_graph_circles.png
TSNE_graph_new.eps
autoenc_lstm.py
crossvalidationscript_lstm.py
crossvalidationscript_test.py
crossvalidationscripttest.py
dcec_xval_pred.py
dcec_xval_test.py
dil_conv_elmo_cuda.py
dilconv_lstm_cuda.py
foo.png
lec_xval.py
lstm_elmo_cuda.py
mystery_backup.py
saved.png
test_list.txt
test_results_0.csv
tsne.py
tsne_dilconv_elmo.py
tsne_generate.py
tsney_list.txt

README.md

CL-AFF-ST

Our codebase for our submission to the AffCon 2019 CL-AFF Shared Task. Still in progress.

File Structure

csv\ contains modified versions of the dataset files provided for the competition, we will not be reuploading the original dataset.

scripts\ contains utility scripts such as our server quick setup script.

lstm_baseline_2label.py is a mutlilabel classification baseline implementation of a naive lstm taking in the default AllenNLP embeddings.

lstm_elmo.py is a baseline implementation of a naive lstm taking ELMo vectors as input and returning both labels as in lstm_baseline_2label.py.

cl_aff_embedders.py contains the custom TextFieldEmbedder that can embed ELMo vectors.

simple_seq2vec.py contains a custom Seq2VecPredictor to predict the social or agency value for a single sentence.

Dependencies

Requires the AllenNLP library and Python 3.6 or above to run.

Next Steps

Unsupervised tasks, continuing characterization of performance, application on other datasets

You can’t perform that action at this time.