This directory (encoder/) contains the code of our two developed methods, as well as the data they use. This data has been extracted within the ../data_extraction/ directory.

In general, everything should be runable with the shell scripts in this folder. They then invoke python scripts found in the subfolder src/, and deal with the data found in the subfolder data/.

prepare_all.sh -> this script prepares the data to be in the right format as is needed by the neural nets. Internally, it runs the script src/preprocess.py. Make sure to set up the right parameters that correspond to the right experiment and method.

The autoencoder can be run with the file run.sh. Make sure you setup the right experiment (crowd experiment or not). Alternatively, you can only run the test cases with the file run_only_test.sh.

To run the EMB model, run the file emb_run.sh, or emb_run_only_test.sh for only evaluating on the test data without training.

In the subfolder src/, there are the implementation of the two baselines that we used in this paper: mfv_baseline.py is the most-frequent-value baseline, whereas naive_bayes_baseline.py contains the implementation of the Naive Bayes method.