Claudinha Data Protection Law (LGPD) Corpus

This dataset contains privacy policies paragraphs in Portuguese. Each paragraph was annotated by an expert annotator, using a guideline (DOI: 10.5281/zenodo.13371432). Two types of notes were made: the category of the Brazilian Data Protection (LGPD) Law that fits the text and the level of compliance with LGPD.

The categories are divided into three blocks: Omission of data required by law (block 1), Data processing (block 2), Unclear language and others (block 3). There are 3 levels of compliance, with level 1 being in compliance with the law, level 2 being partial potential non-compliance, and level 3 being potential total non-compliance.

There are 6341 distinct paragraphs. The corpus has more records (8341 clauses), as there are duplications, since a paragraph can belong to more than one guideline category. Pontetially non-compliant clauses corresponds to 1413 records (21.9%). Below, statistics regarding the number of paragraphs belonging to each category and frequencies of categories in privacy policies.

Category

Number of clauses

Document frequency

Block 1: Omission of data required by law

Access to data
283
61

Anonymization, blocking and deletion
204
46

Automated decision
45
19

Category of processed data
1427
63

Controller identification
107
47

Data correction
154
52

Duration of treatment
234
52

Existence of treatment
142
37

Express consent
176
41

ID and contact DPO
150
46

Non-consent
91
33

Personal data source
471
55

Portability
97
36

Purpose of sharing
119
16

Purpose of treatment
1620
71

Revoke consent
154
50

Right of deletion
173
40

Third party sharing
919
69

Block 2: Data processing

Advertising
215
38

Children data
118
37

Cookies
432
60

Consent by use
339
47

Other consents
73
30

Policy changes
210
61

"Take it or leave it"
50
21

Block 3: Unclear language and others

Generic expressions
244
37

Other unclear clauses
94
26