Parse provider credentials
Heuristichealthcare_credentials_parseBest-effort — fuzzy, approximate, or inferred; check the confidence score.
Parse a healthcare credential string like 'MD, PhD, FACP' into matched degrees and certifications. Multi-pass strategy: delimiter split, period normalize, hyphen-split ('APRN-CNM' → 'APRN', 'CNM'), space-split, license-number strip, roman numeral filter, state-code vs. credential disambiguation (MD/PA/DC exception list), and optional fuzzy matching for typos — applied only to tokens after the first delimiter, so name words in an undelimited leading segment ('Jane Doe MD') are never fuzzy-guessed into credentials; they land in `unknown`. Returns a list of matched tokens each with `type`, `abbr`, `full_name`, and per-token confidence. Processing multiple inputs? Use `healthcare_credentials_parse_batch` to do up to 1000 in one call.
Parameters
textstringrequiredCredential string, e.g. 'MD, PhD, FACP'.
enable_fuzzybooleanoptionalEnable fuzzy matching for close misses. Applies only to tokens after the first delimiter; a leading undelimited segment (often a person's name) must match exactly regardless of this flag — a deliberate guard against fabricating credentials from name words.
Call it
curl -X POST https://api.snipget.ai/v1/healthcare/credentials/parse \
-H "Authorization: Bearer sk_live_..." \
-H "Content-Type: application/json" \
-d '{"text":"MD, PhD, FACC"}'The Batch tab sends up to 1000 inputs in one call via /v1/healthcare/credentials/parse/batch.
FAQ
What does this do for a messy credential string?
It splits something like 'MD, PhD, FACP' into separate structured tokens, each labelled as a degree or certification with its full name, drawing on a maintained credential dataset rather than guessing.
How reliable is the parsing?
It is a best-effort parse with fuzzy matching, so check the confidence and the 'unknown' list. Tokens it cannot place are returned in 'unknown' rather than silently dropped or errored.
What if the string starts with a person's name, like 'Jane Doe MD'?
Leading name words are not fuzzy-guessed into credentials, even with enable_fuzzy left on. Fuzzy matching only applies to tokens after the first delimiter (comma, semicolon, slash, ampersand, 'and'); everything before it must match a known credential abbreviation or observed alias exactly, so 'Jane' and 'Doe' land in 'unknown' while 'MD' still parses. That restriction is deliberate: without it, near-miss matching fabricated credentials from name words ('Doe' looked like the degree DO). Cleanest results come from separating the name with a comma.
Can I turn off the fuzzy matching?
Yes. Set enable_fuzzy to false to require exact matches only. That is stricter and avoids near-miss guesses when you would rather see a token flagged as unknown. The flag only reaches tokens after the first delimiter; it never extends fuzzy matching into the leading name segment described above.
How is this different from expand?
Parse handles a full multi-credential string and returns several tokens plus a summary. Expand handles a single abbreviation. Use parse for a list, expand for one item.