Speech into Text

Using a Custom Dictionary for Names and Technical Terms

Published: October 14, 2026

Every field has words that a general speech model may not recognise. A doctor dictating notes uses drug names and anatomical terms. A developer dictating documentation uses function names and product terms. A teacher uses the names of students, schools, and historical figures. A journalist uses the names of people and places that appear in the news every week. When a speech tool does not know these words, it guesses, and the guess is often wrong. The result is text that needs far more correction than it should.

A custom dictionary solves much of this problem. It gives the tool a list of words, names, or phrases that you use often, so that when it hears something similar, it can choose the correct spelling. This article explains how to use the custom dictionary in Speech into Text, how to choose the right entries, and how to keep the list useful over time.

Why general models struggle with specialist words

Speech recognition models are trained on large collections of everyday language. They learn common words and common phrases very well. Specialist vocabulary, however, appears much less often in that training material, so the model has less evidence about how it sounds and how it is spelled. When you say a rare word, the model chooses the most likely common word that sounds similar, even if that word is not what you meant.

Names are a particular challenge. Many personal names and surnames are uncommon, and some have several accepted spellings. Brand names and product names are often invented words, which means the model has never seen them at all. These are the cases where a custom dictionary helps most.

Opening the custom dictionary

Speech into Text includes a custom dictionary where you can add the words you use most often. You can open it from the tool, add your entries, and save the list. Once saved, the tool uses your entries when it converts your speech into text, so the words you added are more likely to appear correctly in your next session.

Start with a small list. Adding ten or fifteen of the most important terms is more useful than adding a hundred words you rarely use. You can always add more later, once you have a clear sense of which words cause problems.

Choosing good entries

The best entries are the words that appear often in your writing and that the tool gets wrong. Begin by dictating a few paragraphs of your normal work, and note every word that is misrecognised. Those words are your first candidates. Add the correct spelling, and check that the entry matches the way you want the word to appear in your text.

Include the exact capitalisation and punctuation you want, where it matters. A product name written with a specific capital letter or a name with an apostrophe should be entered in that form. If a term has several common spellings, choose the one used in your organisation or publication, and stay consistent.

Phrases as well as single words

You can also add short phrases. A company name made of three words, a technical term that is usually written as a phrase, or a standard expression used in your field can all be entered as a whole. Phrases can be more reliable than single words, because the surrounding words give the tool more context to work with.

Keep phrases short and specific. A very long phrase is less likely to match exactly when you speak, and it is harder to maintain. If a phrase rarely causes trouble, leave it out of the list.

Names of people and places

Personal names are some of the most common sources of error in dictated text. Add the names of the people you write about most often, including colleagues, clients, family members, and public figures you cite. For places, add the names of your city, your organisation's address, and any locations that appear frequently in your work.

Be careful with names of private individuals. Only add the information you need for your own writing, and keep the list private. Remember that the dictionary is stored in your browser on your device, so it is available only on that device.

Technical and specialist terms

Specialist fields benefit most from a custom dictionary. In medicine, you might add drug names, procedures, and anatomical terms. In law, you might add the names of statutes, court terms, and legal phrases. In software, you might add function names, programming terms, and product names. In science, you might add chemical names, species names, and units. Create a list for your own field and update it as you learn new terms.

When a term has an unusual pronunciation, add a spelling that matches how you usually say it, and check that the result is what you intended. The goal is not to change the meaning of your speech but to help the tool produce the correct written form.

Keeping the list useful

A custom dictionary is only useful if it stays accurate. Review the list every few weeks. Remove entries you no longer use, correct any spelling that has changed, and add new terms as your work changes. A short, well-maintained list will serve you better than a long, outdated one.

Keep a copy of your list somewhere safe. If you clear your browser data or move to a new device, the dictionary may not be available. A simple text file with your entries makes it easy to rebuild the list quickly.

Checking results after dictation

The custom dictionary reduces errors, but it does not remove the need to check your text. After each session, read the text and look for any words that are still wrong. If you find a new error, add the correct word to the dictionary so that the mistake is less likely to happen again. Over time, this simple cycle of dictating, checking, and updating produces a list that matches your work very closely.

Sharing a list with a team

If several people in a team use the same terms, you can share a text list of the agreed spellings. Each person can then add the entries to their own dictionary. This helps keep product names, job titles, and client names consistent across documents. Make sure the list does not include confidential information that should not be shared.

Summary

A custom dictionary helps speech recognition get the names and specialist words in your work right. Start with the terms that cause the most errors, add phrases where they help, and keep the list short and up to date. Check your text after each session, and add any new mistakes to the dictionary. With regular maintenance, the dictionary becomes one of the most useful parts of your voice typing routine.

To set up your own list, open the speech to text tool and add the first five terms you use most often.