Custom vocabulary sounds like a small feature until you hear a name, medication, or case term get mangled three times in a row.
OPI calls depend on small words with large consequences: “Eliquis,” “subrogation,” “Medicaid,” “arraignment,” “aneurysm,” “deductible.” If a live transcript or speech tool keeps guessing the wrong term, the interpreter has to spend extra attention correcting the screen instead of listening.
Term mappings help the tool meet the call halfway. You give it the terms you expect, and it gives those terms more weight when it listens or translates.
For healthcare vocabulary discipline, the NCIHC standards of practice are a useful guardrail.
A custom term list is useful only when it matches the words your callers actually say.
What a term mapping does
A term mapping tells your tool: when you hear this source term, prefer this spelling, rendering, or target-language equivalent.
TIP
Add terms after real misses, not after abstract study sessions. Your glossary should reflect your calls.
That can mean:
- A medication name and its correct spelling
- A provider name or facility name
- A legal term with an accepted equivalent
- An agency acronym
- A product name that speech recognition keeps misreading
- A phrase that should stay untranslated
In Interpreter, you can load Important Words and term pairs before a session. Other tools may call the same idea a glossary, vocabulary boost, custom dictionary, or forced replacement. The feature name changes. The workflow stays similar.
You give the system a short list of terms that matter for this call.
Keep the list short
Many interpreters overbuild the list. They add every word from a prep document, every possible medication, and every acronym they might hear. Then the tool has too much context and the useful terms stop standing out.
For live OPI, a smaller list works better:
- Names
- Uncommon medication or procedure terms
- Case-specific acronyms
- Brand names
- Terms with known mistranscriptions
- Words with high risk if confused
If you are interpreting a cardiology call, you do not need the whole cardiology textbook. You may need “metoprolol succinate,” “atrial fibrillation,” “echocardiogram,” and the cardiologist’s name.
Think of custom vocabulary as a call kit, not a dictionary.
Use mappings for recognition, not decision-making
A term mapping can improve what appears on screen. It cannot decide what the speaker meant.
If a patient says something that sounds like “insulin,” the tool may show “insulin” because you loaded it. That helps only if the speaker said insulin. If the audio was unclear, you still need to ask for repetition.
The interpreter stays responsible for the message. Treat the screen as a support, much like Quick Lookup or notes. Use it to reduce friction, then apply your training.
Medical and legal calls make that responsibility sharper. Similar terms can point to different realities:
| Heard term | Possible confusion |
|---|---|
| ileum | ilium |
| hypertension | hypotension |
| appeal | repeal |
| premium | prepayment |
| discharge | discharge summary, discharge from care, discharge from debt |
Mappings can help the tool prefer the term you expect. They cannot remove the need to clarify.
Build reusable sets by domain
You do not need to rebuild from scratch before every shift. Create small reusable sets:
Medical intake: primary care provider, date of birth, allergies, current medications, pharmacy, pain scale.
Pharmacy: refill, prior authorization, dosage, generic, brand name, contraindication, side effect.
Insurance: deductible, copay, coinsurance, claim number, member ID, appeal, network.
Legal: arraignment, affidavit, deposition, plea, continuance, probation, warrant.
Utilities: shutoff notice, account number, past due, payment arrangement, service address.
Each set should stay lean. You can add call-specific terms on top when the portal, patient, or client gives you clues.
For broader prep, use a more permanent glossary system. Our interpreter glossary bank guide explains how to store terms across months of work without turning your live-call setup into clutter.
Map both directions when the call is bilingual
OPI does not move in one direction. The provider may speak English. The patient may answer in Spanish, Arabic, Mandarin, Uzbek, or another language. Then the provider may quote the patient’s words back.
If your tool supports it, add both directions:
- English source term to target-language equivalent
- Target-language term or spelling variant to English
- Common abbreviation to full term
- Brand name to generic name when you know the call requires both
This helps with live transcripts and with post-call review. It also reduces one common frustration: the tool catches the provider’s term but misses the patient’s version of the same idea.
Do not invent equivalents. Use accepted professional terminology, your agency guidance, and your own verified glossary.
Avoid patient-specific data
A custom vocabulary list can create a privacy problem if you fill it with names, dates of birth, claim numbers, addresses, or medical record numbers and then store it after the call.
Use the minimum. If you need a name during the call, remove it afterward unless your agency-approved tool handles session cleanup for you. Do not paste a full referral, chart note, or insurance letter into a vocabulary field unless the tool has approval for that data.
Healthcare interpreters should pair vocabulary tools with privacy habits. Start with HIPAA for interpreters if your workflow includes patient information.
Review what the tool got wrong
After a hard call, look at the terms the tool missed or distorted. Add the reusable ones to your glossary. Remove one-off details. If the same term keeps failing, create a mapping before the next related call.
That review should take five minutes, not an hour.
Term mappings work best when you treat them as preparation for the next likely mistake. Load the few words that would cost you the most attention if the system missed them. Then let the tool do its job quietly while you do yours.
For many interpreters, that is the real value: fewer tiny interruptions, fewer spelling guesses, and more attention left for the human conversation.
Related reading: