Custom Vocabulary

Improve transcription accuracy with domain-specific terms, names, and acronyms.

What is Custom Vocabulary?

Custom vocabulary lets you define terms that the speech recognition system should prioritize during interviews. This is especially useful for:

• Expert names with unusual spellings (e.g., "Mkrtchyan-Voznesenskaya")

• Company names that may be misheard (e.g., "NovoCure Therapeutics")

• Industry acronyms (e.g., "PD-L1", "MSI-H", "TMB")

• Technical jargon specific to your research domain

• Product names or proprietary terms

Without custom vocabulary, the speech recognition may transcribe unfamiliar terms phonetically, leading to errors like "Novo Cure" instead of "NovoCure" or "P-D-L one" instead of "PD-L1".

How It Works

Custom vocabulary improves transcription at two stages:

1. Real-time recognition — Terms are passed to the speech recognition engine during the interview as keyword hints, improving accuracy as the conversation happens

2. Post-call cleanup — After the interview completes, an AI model reviews the transcript against your vocabulary and corrects any remaining errors

Both account-level vocabulary and AI-generated terms (based on interview context) are combined automatically for each interview.

Managing Vocabulary

Navigate to Vocabulary Settings

1. Go to Vocabulary in the sidebar

2. You'll see your categories and terms, along with usage stats

Categories

Organize terms into categories for easier management:

Action	Steps
Create category	Click Add Category, enter a name
Rename category	Click the edit icon next to the category name
Delete category	Click the delete icon (removes all terms in the category)

You can have up to 20 categories per account.

Adding Terms

Manual Entry

1. Expand a category

2. Type the term in the input field at the bottom

3. Click Add

Import from CSV

1. Click Import CSV

2. Choose an import mode:

   • From CSV categories — CSV includes a category column; terms are sorted into matching categories (new categories are created automatically)

   • Into existing category — All terms go into a selected category

   • Create new category — Create a category and import all terms into it

3. Upload or paste your CSV content

4. Click Import

CSV format for multi-category import:

category,term
Biomarkers,PD-L1
Biomarkers,TMB
Companies,NovoCure Therapeutics

CSV format for single-category import:

term
PD-L1
TMB
MSI-H

AI Suggestions

Get vocabulary suggestions based on your interview context or expert profiles:

1. Click Get AI Suggestions

2. Choose a source:

   • From interview — Generates terms based on interview title, questions, and expert data

   • From expert — Generates terms based on expert name, role, and employment history

3. Select a target category for the suggested terms

4. Review suggestions — terms already in your account are marked

5. Select the terms you want to add

6. Click Add Selected

Term Limits

Limit	Value
Maximum terms per account	1,000
Maximum words per term	6
Maximum categories	20
Maximum terms per category	500

The 6-word limit is enforced by the speech recognition engine. Multi-word terms like "Adaptive Accelerated Protocol Trials" count as 4 words.

Per-Interview Customization

When creating or editing an interview, you can customize which vocabulary terms are used:

1. In the interview form, find the Vocabulary section

2. The total term count shows account terms plus any interview-specific additions

3. Click Customize to open the vocabulary modal

4. From the modal you can:

   • Search across all categories to find specific terms

   • Remove account terms you don't want for this interview

   • Add new terms specific to this interview

5. Changes only affect this interview — your account vocabulary stays unchanged

How Terms Are Combined

For each interview, the final vocabulary is computed by combining:

1. Account vocabulary — All terms from your account settings

2. Interview overrides — Terms added or removed for this specific interview

3. AI-generated terms — Automatically generated based on interview context (questions, expert data)

The system deduplicates and truncates to stay within the 1,000 term limit.

Export

Export your vocabulary as a CSV file:

1. Click Export on the Vocabulary page

2. A CSV file downloads with category and term columns

3. Use this to back up your vocabulary or transfer it to another account

Best Practices

Term Selection

• Focus on terms the speech engine is likely to get wrong — unusual names, acronyms, and specialized jargon

• Include both the acronym and its expansion if both may be spoken (e.g., "AAPT" and "Adaptive Accelerated Protocol Trials")

• Add expert names before scheduling interviews for the best recognition

Category Organization

• Group terms by domain or project for easier management

• Use categories like "Expert Names", "Companies", "Medical Terms", "Acronyms"

• Keep category names descriptive — they're visible when customizing per-interview

Ongoing Maintenance

• Review transcripts after interviews to identify terms that were still misrecognized

• Add commonly misheard terms to your vocabulary

• Use AI suggestions when starting research in a new domain to quickly build a relevant vocabulary

Troubleshooting

Terms not improving transcription

• Verify the term is included in the interview's computed vocabulary (check the Vocabulary section on the interview detail page)

• Terms must be added before the interview starts — they cannot retroactively fix completed transcripts

• Very short terms (1-2 characters) may not be effective as keyword hints

AI suggestions returning empty

• Ensure the interview has a title and questions, or the expert has a name and employment history

• The AI needs context to generate relevant suggestions

Import not working

• Check that your CSV uses the correct column headers (category, term)

• For single-category import, select the target category before importing

• Terms that exceed the 6-word limit are skipped during import