Quantitative Testing with Concept Activation Vectors (TCAV) in NLP

Data preparation

We use Lenta.Ru dataset in our experiments.

Create data_path folder and put lenta-ru-news.csv file in it
Choose labels to experiment with and split the data into train and test by running lenta_dataset.ipynb
Finally, run python data_prep.py -dd data_path. This command will save full.csv, train.csv, eval.csv and vocabs.txt files to data_path

Modify build_model function if you want to change an architecture of the model
Create experiment_path folder and put experiments/config.yaml in it
Modify hyperparameters inside experiment_path/config.yaml
Run python train.py -dd data_path -md experiment_path. This command will train the model and save checkpoints to experiment_path

Choose words you would like to experiments with. For example, Москва, ООН, Жириновский will be a good choice.
Run python collect_concepts.py -dd data_path -md experiment_path --ngrams 3. This command will generate multiple files:

concepts.pkl -- for each concept (e.g. Москва) we search for sentences where this word occurs. Then we retrieve ngrams of size n from this sentence (e.g. лето в Москва, Москва слезам не верит) and call it concepts. Also we collect some random samples from the data for each concept.
cav_bottlenecks.pkl -- we convert concept texts into hidden representations of the model from experiment_path folder
cavs.pkl -- Hyperplanes for each concept received by fitting Logistic Regression on concept/non-concept data. LR is trained on hidden representation of the data.
grads.pkl -- directional derivatives (see the paper for more details)

Run python calculate_tcav.py -dd data_path. This command will save scores.pkl file. In this file you can find TCAV scores for each concept against all labels.

Run TCAV.ipynb to compare results.