How can I create, edit or delete a dictionary?
Manage Dictionaries for Optimized Transcription in Deep Live Hub
What does a dictionary?
When ever the ASR model is low in confidence during a transcription, it takes the dictionary as additional source and evaluates if a dictionary term might be a better fit for the transcription. If you provide additional "sounds like" entries to the dictionary term, the transcription gets even more accurate.
How to Create a Dictionary:
-
Log in to Your Account:
- Navigate to the "Dictionaries & Glossaries" section.
-
Create a New Dictionary:
- Click on "Create Dictionary".
- Enter a name for your dictionary in the name field and press "Save".
How to Edit a Dictionary:
-
Add Words Manually:
-
Click Add Word to add a new entry.
-
In the Word field, enter how the word should appear in the transcript.
-
In the Sounds Like field, enter the phonetic spelling of how the word is pronounced (see description below).
-
Press Enter to save the word.
-
- Upload a List of Words (CSV Import):
-
- Create a dictionary by following the steps outlined above.
-
Open the Import Option:
-
Click the three dots (⋯) next to the Add Word button.
-
Select Import CSV from the dropdown menu.
-
-
Upload a CSV File:
-
Choose a CSV file from your device.
-
Ensure the file meets the required formatting guidelines before uploading.
-
-
Add Words:
- All words from the CSV file will be automatically imported and added to the dictionary.
- Delete Words:
- Go to the "Dictionaries & Glossaries" overview page and select the dictionary you want to edit.
- Check the boxes next to the words you wish to delete and click "Delete Selected Words".
Info: CSV Formatting Requirements
A CSV (Comma-Separated Values) file is a simple text file where each entry appears on its own line. This format allows you to quickly add multiple words to your dictionary at once and can be created or edited using common tools such as Excel or Word.
Updated Format with “Sounds Like” Support
To support the Sounds Like feature, your CSV file must follow this structure:
-
Start with the dictionary word
-
Separate the word from its phonetic entries using a semicolon (;)
-
Add one or more Sounds Like entries, separated by commas (,)
Example:
CEO;C.E.O.,see-E-oh
SOC 2;Sock2,Sock-2
This ensures that both the word and its pronunciation variations are imported correctly in a single step.
How to use "Sounds Like" entries:
Additional you can also add "Sounds Like" entries for each term, increasing the transcription quality.The idea behind "Sounds Like" entries is that you provide the ASR model with additional information on how this dictionary item could also be transcribed phonetically, to improve the detection and transcription of names or special terms. The 'Sounds Like' entry should not be written in the phonetic alphabet; it should be written in regular language to provide an alternative way in which the ASR could understand it.
For example, you might want the word 'CEO' to always be written this way: CEO
-
- The ASR model usually understands "See E Oh" and automatically transcribes it as "C.E.O."
- So without a "Sounds Like" entry, it transcribes "C.E.O." as an regular abbreviation with dots.
- Now create a dictionary entry for "CEO" with the "Sounds Like" parameters "See E Oh" and "C.E.O" as Sounds like variants.
- The ASR model will now use these phonetic alternatives as additional reassurance for its transcription. The word 'CEO' is now always written correctly, without dots.
This function is particularly important when working with names from other languages. An ASR model is always trained on specific language(s); therefore, names and entities from other languages can be transcribed incorrectly. Using "Sounds Like", however, gives you the ability to provide the correct transcription and tell the model what phonetic sound to look for when replacing it with this dictionary item.
Note:
Make sure you don't exceed six words for each dictionary item and its 'Sounds Like' entries combined! The system will ignore them.
How to Delete a Dictionary:
-
Go to the Overview Page:
- In the "Dictionaries & Glossaries" overview page, find the dictionary you want to delete.
-
Delete the Dictionary:
- Click the three-dot symbol next to the dictionary name.
- Select the "Delete" button to remove the dictionary.
How to Activate or Deactivate a Dictionary in Your Workflow:
-
Log in and Go to the Workflow Editing Page:
- Open the live workflow where you want to use the dictionary.
-
Activate the Dictionary:
- Go to the Process tab and select your dictionary from the dropdown menu in the middle column.
- Click "Save" to activate the dictionary for this workflow.
-
Deactivate the Dictionary:
- To remove the dictionary, simply select the default "Dictionary" option from the dropdown and press "Save".
One Dictionary Per Workflow:
You can only activate one dictionary per workflow.
Word Limit:
Avoid adding more than 1000 words to a dictionary, as larger dictionaries may increase the error rate of the speech-to-text recognition system.