Redaction

elevateai's redaction feature automatically detects and replaces personally identifiable information in transcripts — helping you meet compliance requirements and protect customer data at scale redaction is configured per token and requires user opt in to get started, navigate to the manage tokens page and click the redactions button to select which entity types to redact and whether redaction should apply to post call, real time, or both how it works when redaction is enabled for a token, elevateai scans each transcript for the entity types you have selected detected pii is replaced with a placeholder in the format \<entity type> before the transcript is stored or returned example "my social security number is \<us ssn> and you can reach me at \<phone number> " details on each redacted segment — including timing, entity type, and confidence score — are returned alongside the transcript in the redactionsegments array ui for editing redaction options configuring redaction redaction settings are managed at the token level from the manage tokens page setting description entity types select the pii categories to detect and redact a full list of supported entities is provided below post call apply redaction to transcripts processed via the elevateai post call api real time apply redaction to transcripts streamed via the elevateai real time api (wss) redaction is only on the final transcript post call api for interactions declared and processed through the elevateai post call api, redaction is applied automatically when it is enabled for the token used to declare the interaction when retrieving a phrase by phrase or punctuated transcript for a redacted interaction, the response will include a redactionsegments array containing details on each detected pii instance sample response — redactionsegments { "redactionsegments" \[ { "starttimeoffset" 1467720, "endtimeoffset" 1468030, "result" "us ssn", "score" 0 98745 }, { "starttimeoffset" 2031100, "endtimeoffset" 2031850, "result" "phone number", "score" 0 95312 } ] } redactionsegments fields field type description starttimeoffset integer start time of the redacted segment in milliseconds endtimeoffset integer end time of the redacted segment in milliseconds result string the entity type that was detected and redacted (e g , "us ssn" , "phone number" ) score number confidence score for the detection, from 0 0 to 1 0 note all source audio data is deleted immediately upon processing please refer to elevateai's data retention & security https //docs elevateai com/overview#mywqd guidelines for additional details real time api (wss) for sessions processed through the elevateai real time api, redaction is applied based on the configuration set for the token on the manage tokens page final transcript redaction — redacts pii in the completed transcript returned at the close of a session supported entity types nlp based entities detected using natural language processing available for english and spanish only entity type description name names of individuals address us addresses global entities detected using pattern matching available for all supported languages entity type description date of birth date of birth credit card credit card numbers (visa, mastercard, amex, and others) email address email addresses healthcare number medical license numbers phone number phone numbers in various formats goverment issues id social security numbers, drivers license numbers, us passport numbers account number us bank account numbers language support elevateai redaction supports 2 languages language code language nlp entities regional entities en , en us english yes us, uk es , es 419 spanish yes es // // using redaction 1\ manage at the api token level redaction configuration options are attached to specific api tokens, allowing token level customization for example, your customer support token might include different substitutions than your sales token 2\ access the phrase editor log into the elevateai portal < app elevateai com > your security level within your account determines your access to the boosted phrases list access level by role role access level owner full access (view, edit, create) admin full access (view, edit, create) api user view only access explore no access navigate to manage api tokens locate the desired token and select edit redaction to open the pii redaction configuration only account owners or admin can edit a find & replace pair 3\ select a language choose the language you wish to edit from the dropdown each api token supports one redaction configuration per language 4\ select redaction configuration select to turn redaction on for either real time(final) or post call, or both select which entities to enable redaction on you can use select all to select all 5\ disabling redaction to transcribe without redaction , use a token that does not have an redaction enabled to remove the redaction feature completely, edit the configuration to disable all redaction elements need more help? need more help? contact the elevateai support team mailto\ support\@elevateai com