Representation Notebook

 

Overview

In this representation notebook, we are evaluating ner.dl model on representation tests. The goal of representation testing is to determine if a given dataset represents a specific population accurately or if it contains biases that could negatively impact the results of any analysis conducted on it.

Open in Collab

Category Hub Task Open In Colab
Representation John Snow Labs NER Open In Colab

Config Used

tests:     
  defaults:
    min_pass_rate: 0.55
  representation:
    min_religion_name_representation_count:
      min_count:
        christian: 10
        muslim: 5
        hindu: 15

    min_label_representation_proportion:
      min_proportion:
          O: 0.5
          LOC: 0.2

Supported Tests

  • min_gender_representation_count: Determine if any gender(male, female or unknown) has less than the desired minimum representation count.

  • min_gender_representation_proportion: Determine if any gender(male, female or unknown) has less than the desired minimum representation proportion.

  • min_ethnicity_name_representation_count: Determine if any ethnicity(black, asian, white, native_american, hispanic or inter_racial) has less than the desired minimum representation count.

  • min_ethnicity_name_representation_proportion: Determine if any ethnicity(black, asian, white, native_american, hispanic or inter_racial) has less than the desired minimum representation proportion.

  • min_label_representation_count: Determine if any label(O, LOC, PER, MISC or ORG) has less than the desired minimum representation count.

  • min_label_representation_proportion: Determine if any label(O, LOC, PER, MISC or ORG) has less than the desired minimum representation proportion.

  • min_religion_name_representation_count: Determine if any religion(muslim, hindu, sikh, christian, jain, buddhist or parsi) has less than the desired minimum representation count.

  • min_religion_name_representation_proportion: Determine if any religion(muslim, hindu, sikh, christian, jain, buddhist or parsi) has less than the desired minimum representation proportion.

  • min_country_economic_representation_count: Determine if any country(high_income, low_income, lower_middle_income or upper_middle_income) has less than the desired minimum representation count.

  • min_country_economic_representation_proportion:Determine if any country(high_income, low_income, lower_middle_income or upper_middle_income) has less than the desired minimum representation proportion.