Safer data opens the door to AI adop­tion

Is your orga­ni­za­tion con­sid­er­ing using sen­si­tive free-text data? Are you unsure whether your cur­rent free-text pro­cess­ing prac­tices meet the require­ments?

In many orga­ni­za­tions, a sig­nif­i­cant amount of crit­i­cal infor­ma­tion still exists in free-text for­mat, such as patient tran­scripts, social care reports, legal deci­sions, and cus­tomer feed­back. Free-text data usu­al­ly con­tains both direct iden­ti­fiers (such as names, address­es, and iden­ti­fi­ca­tion num­bers) and indi­rect iden­ti­fiers (such as pro­fes­sions, loca­tions, and event dates). In many cas­es, even a com­bi­na­tion of indi­rect iden­ti­fiers, such as occu­pa­tion and place of res­i­dence, may enable the iden­ti­fi­ca­tion of an indi­vid­ual in the text.

How­ev­er, man­u­al anonymiza­tion involv­ing human inter­ven­tion is slow, cost­ly, and in many cas­es impos­si­ble when deal­ing with hun­dreds of mil­lions of texts. For datasets of this scale, anonymiza­tion using large lan­guage mod­els is also too expen­sive and slow.

Tieto is con­tin­u­ous­ly devel­op­ing and expand­ing its capa­bil­i­ties for pro­cess­ing var­i­ous types of free-text data.

Read more Hun­dreds of mil­lions of rows of sen­si­tive text: how can per­son­al infor­ma­tion be removed?

More infor­ma­tion

Tieto News­desk

news@tieto.com

Source (text and image): Tieto