Use cases

Find personal data in free-text columns

Before a dataset is shared, check each field for names, contact details or health information.

Ask for an invite

Try it on this example

Example · Notes column of a delivery support export: a relative, a health condition and a phone number

Text of one cell from a free-text column

Rang back re missed delivery on order 48213. Customer said her husband Tom usually takes the parcels in but he has been in hospital with his diabetes since Monday, so please leave with no. 14 next door. Call her on 07700 900123 after 6 if the driver cannot get an answer. Dan
  1. Does the text contain information about a person who could be identified from it?Yes99%
  2. What is the most sensitive kind of personal data in the text?Health100%
  3. Does the text give details about someone other than the person the record is about?Yes98%
  4. Does the text give details about a child?No91%
  5. Can this text be shared as it is, under our sharing rule?No92%

These are real answers stored from one run on this example.

The prism behind it

Find personal data in free-text columns5 questions

Fields

  • Text of one cell from a free-text column

Context

The text is one cell from a free-text column, such as agent notes, comments or a description, in a dataset our team is about to share with another team, a supplier or an analytics tool. The structured columns have already been removed or masked in code; this check is for what people typed into the free text. The answers flag the cells a person must redact or hold back. They do not say where in the text the details are, and nothing is shared or deleted on these answers alone. What counts: - Personal data is any information about a person who can be identified from it, directly or with other details: a name, a phone number, an email or postal address, an account, order or ID number tied to a person, or details that single out one person, such as "the woman at number 14". - Special category data is information about health, disability, ethnic origin, religion, political views, trade union membership, sex life or sexual orientation, genetic or biometric data, and information about criminal offences. - Financial data is a card number, a bank account number, income, debts or details of payments tied to a person. Our sharing rule: a cell may be shared as it is only when it contains no personal data about anyone except the first name or staff code of one of our own staff, and our own business details.

Questions

  1. Does the text contain information about a person who could be identified from it? Yes / No

    Use the definition of personal data in the context. Our own staff's first names or staff codes, and our own business details, do not count. Yes: The text contains personal data about at least one customer, relative, neighbour or other person. No: The text contains no personal data about anyone, or only staff first names, staff codes or our own business details.

  2. What is the most sensitive kind of personal data in the text? Choice

    Pick the most sensitive kind present, in this order from most to least: Health, Other special category, Financial, Government ID, Contact details, Name or identifier. Pick None when the text contains no personal data under the rule in the context.

    • Health A physical or mental health condition, illness, injury, disability, medication, pregnancy or a stay in hospital.
    • Other special category Ethnic origin, religion, political views, trade union membership, sex life or sexual orientation, or a criminal offence, with no health details.
    • Financial A card or bank account number, income, debts, benefits or payment difficulties of a person.
    • Government ID A national insurance, social security, passport, driving licence or tax number.
    • Contact details A phone number, email address or postal address of a person.
    • Name or identifier A name, username, account or order number tied to a person, or a description that singles someone out, with nothing more sensitive.
    • None No personal data, or only staff first names, staff codes or our own business details.

  3. Does the text give details about someone other than the person the record is about? Yes / No

    Count a relative, partner, neighbour, carer, colleague or other person described or named in the text, such as "her husband" or "the man next door". Our own staff do not count. Yes: The text gives details about another person besides the one the record is about. No: The text is only about the person the record is about, or about no one.

  4. Does the text give details about a child? Yes / No

    Count a son, daughter or other person the text says or shows is under 18, such as a pupil, a baby or a teenager. Yes: The text gives details about a child. No: The text mentions no child.

  5. Can this text be shared as it is, under our sharing rule? Yes / No

    Use the sharing rule in the context. Yes: The text contains nothing the rule keeps back, and can go out as it is. No: The text contains at least one detail the rule keeps back, so a person should redact or hold it back.

Lens columns

contains_personal_data, contains_personal_data_probability, most_sensitive_type, most_sensitive_type_probability, third_party_mentioned, third_party_mentioned_probability, child_mentioned, child_mentioned_probability, safe_to_share, safe_to_share_probability

Run it on your own text

Add this prism in the app, change any question, and test it on a file of your own.

Ask for an invite