Tested with Python 3.7.4.
Every dependency can be found in the Requirements file.
pip install -r requirements.txt
glove_cli.py <embedding_path> <semcat_dir> [-h] [-dense_file] [--lines_to_read LINES_TO_READ] [--mcrae_dir MCRAE_DIR] [-mcrae_words_only] [--weights_dir WEIGHTS_DIR] [-save_weights] [-load_weights] [--calculate CALCULATE] [--calculation_args [CALCULATION_ARGS [CALCULATION_ARGS ...]]]
- embedding_path - Path to the Glove embedding file (f.e. "glove/glove.6B.300d.txt")
- semcat_dir - Path to the SemCat categories directory (f.e. "semcat/Categories")
- dense_file - If embedding_path points to a dense embedding file, mark it with this parameter
- lines_to_read - Maximum vectors to read. Default -1 (All vector)
- mcrae_dir - McRae directory
- mcrae_words_only - Use McRae words only
- weights_dir - A path where the weights going to be saved to or read from (f.e. "weights/"), Default "out/"
- save_weights - Save weights to weights_dir
- load_weights - Load weights from weights_dir
- calculate - Calculation method [score|decomp]
- calculation_args - List of arguments for calculation:
- score:
- [int ] <- Lamda value which > 0. Optional: Default 1.
- decomp:
- [str ] <- The word to decompose
- [int ] <- The top X category. Optional: Default 20.
- [bool ] <- Save result into file. Optional: Default False.
- [str ] <- The word to decompose
- score:
Example:
python interpret_cli.py "data/glove/glove.6B.300d.txt" "data/semcat/Categories" -dense_file --lines_to_read=50000 -load_weights --calculate=decomp --calculation_args barrel 10 True