| 123456789101112131415161718192021222324252627282930313233343536373839404142434445464748 |
- from datetime import datetime
- from collections import OrderedDict
- import pathlib
- """
- Configuration file for Text Guide.
- How to use:
- Define folder containing data file, filename of the data file, name of the column in the data file containing
- unstructured textual data, select the truncation method and the name of the file (dictionary) containing
- pairs of "key: value" where key == important token, and value == value of feature importance. Select the sorting
- method for the important token dictionary.
- """
- class ExperimentConfig(OrderedDict):
- def __init__(self):
- super().__init__()
- self.start_time = datetime.utcnow()
- self.config_name = f"{pathlib.Path(__file__).stem}"
- self.df = None
- # specify files and names
- self.data_folder = 'data'
- self.filename = "dmoz_100_instances.ftr"
- self.text_column = "Text"
- self.nrows = 7 # the number of instances to convert. The Whole file dmoz_100_instances.ftr has 100 instances.
- self.truncation_method = 'text_guide' # or 'beginning'
- self.feature_importance_file = f"dmoz_30_1500_sITFL.p" # the attached file was obtained from a BoW model
- # and a gradient boosting classifier. For different data, different file is needed.
- self.feature_importance_sort = 'descending'
- # specify other Text Guide parameters
- self.desired_length = 300 # the desired length of the new text instance.
- self.one_side_neighbours = 3 # number of tokens surrounding the important token to be used for creating the
- # pseudo sentence
- self.beg_part = 0.1 # the part of the beginning of the original text instance to be used by Text Guide to
- # create the new text instance
- self.end_part = 0.2 # the part of the ending of the original text instance to be used by Text Guide to
- # create the new text instance
- self.over_length = 1 # if set to 1, Text Guide will be used for all tex instances. Text Guide will be used
- # only for instances with length greater than desired_length*over_length tokens.
- self.number_of_important_token_occurrences = 1 # number of important token occurrences to be used by
- # Text Guide to create pseudo sentences.
- self.fill_up_to_limit = True # some text instances created by Text Guide will be shorter than the
- # desired_length. Set to "True" if you wish to fill the remaining space by initial tokens of the original text
- # instance
- self.cpu_threads = 14 # number of cpu threads to be used by Dask when computing Text Guide instances
|
