In this repo, you can find the data from our ACL 2021 paper "HateCheck: Functional Tests for Hate Speech Detection Models".

• "test_suite_cases.csv" contains the full test suite (3,728 cases in 29 functional tests).
• "test_suite_annotations.csv" provides detailed annotation outcomes for each case in the test suite.
• The corresponding "all_" files cover all 3,901 cases that were initially generated, from which 173 were excluded from the test suite due to fewer than four out five annotators agreeing with our gold standard label.
• "template_placeholders.csv" contains the tokens that the placeholders in the case templates are replaced with for generating the test cases.

functionality The shorthand for the functionality tested by the test case.

case_id The unique ID of the test case (assigned to each of the 3,901 cases we initially generated)

test_case The text of the test case.

label_gold The gold standard label (hateful/non-hateful) of the test case. All test cases within a given functionality have the same gold standard label.

target_ident Where applicable, the protected group targeted or referenced by the test case. We cover seven protected groups in the test suite: women, trans people, gay people, black people, disabled people, Muslims and immigrants.

direction For hateful cases, the binary secondary label indicating whether they are directed at an individual as part of a protected group or aimed at the group in general.

focus_words Where applicable, the key word or phrase in a given test case (e.g. "cut their throats").

focus_lemma Where applicable, the corresponding lemma (e.g. "cut sb. throat").

ref_case_id For hateful cases, where applicable, the ID of the simpler hateful case which was perturbed to generate them. For non-hateful cases, where applicable, the ID of the hateful case which is contrasted.

ref_templ_id The equivalent, but for template IDs.

templ_id The unique ID of the template from which the test case was generated (assigned to each of the 866 cases and templates from which we generated the 3,901 initial cases).

functionality, case_id, templ_id, test_case, label_gold See above.

label_[1:10] The label provided for the test case by a given annotator. We recruited and trained a team of ten annotators. Each test case was annotated by exactly five annotators.

count_label_h The number of annotators who labeled a given test case as hateful.

count_label_nh The number of annotators who labeled a given test case as non-hateful.

label_annot_maj The majority label.