Custom Vocabulary for Transcription

CXone Mpower Transcription provides a continuous stream of transcription in real time. Transcription uses an ASR Automatic Speech Recognition. Allows contacts to respond to prompts by speaking, pressing phone keys, or both. model that has been trained on contact center conversations. The conversations used to train the model span a wide range of industries. However, every organization uses words that are unique, or that are used in unique contexts. This can affect the accuracy of transcription results.

You can tune the model that Transcription uses. This allows you to tune the model so it recognizes terminology that's unique to your organization or that has a unique context in your line of business.

Customization Options

You can use the following customizations with CXone Mpower Transcription: 

• Boost words: Increase the likelihood that Transcription chooses a particular word when transcribing.
• Find and replace words: You can have Transcription replace certain words in the final transcription results with other words or phrases. A maximum of 200 find and replace configurations are supported across all enabled languages.

Transcription supports a maximum of 50 boosts and 200 find and replace word configurations across all enabled languages. To use these customizations efficiently, prioritize using them for business-specific terminology that's not likely to vary between languages. This includes product names, locations, or other standardized terms.

Before Customizing Vocabulary

You can improve transcription results by customizing the vocabulary. This helps the Transcription engine better recognize important words and phrases that are common in your industry. When the engine can recognize these words more reliably, it makes fewer mistakes and the transcripts need less editing.

Vocabulary customizations are a valuable tool, but they should be used with care. Choosing the wrong words or using the customization options for the wrong reasons can cause problems rather than solve them. Before you make changes, this section will guide you through what to think about and how to avoid common issues.

Test Transcription and Determine Error Types

Only customize vocabulary based on real problems you've seen in actual transcription results. When a new version of the Transcription engine is released, your existing vocabulary configurations might need updated. Before you make changes, review transcription results from the new version to determine what kind of changes you need to make.

As you test Transcription, keep track of any errors you notice. Group the errors by type. The type of error will help you decide whether a customization can fix it. The error types are described in the following table.

Error Type	Details
Words a native speaker may not know	Words could be proprietary, rare, foreign, or industry-specific terms. There are two possibilities with this type of error:  Consistent Failure: Transcription always gets the word wrong in the same way. For example, the Japanese word hibari is frequently used by English-speaking contacts. Transcription consistently transcribes it as he body. Correct it by defining a word find and replace. Inconsistent Failure: Transcription gets the word wrong in different ways. For example, the Japanese word hibariis sometimes transcribed in English as he body and sometimes as the body. Correct it by defining a find and replace configuration. If this is not a high-value word, consider boosting the correct word instead.
Company and product names, abbreviations, and acronyms	These are words that Transcription is unlikely to recognize or that may mean something else in a different context. They may also be words that are not transcribed with correct capitalization. Correct these errors by defining a find and replace configuration. For example, if NVIDIA is transcribed as nvidia, a find and replace configuration will instruct Transcription to replace its lower case version with the upper case version.
Standard Vocabulary	Words that the average native speaker of the language recognizes but that are failing to transcribe correctly. Important: Correcting transcription failures of standard vocabulary words is not recommended. Attempting to correct failures of this type can cause more errors than it fixes.

Choose Customizations Strategically

It's important to choose which words to configure customizations for. Using too many customizations can lead to an increase of inaccurate results. Additionally, the numbers of each customization type are limited, so avoid using customizations on low-value words.

For example, if you boost too many words, Transcription might overuse them. This can cause words to be used in the wrong context, which makes the transcript less accurate. Similarly, using find and replace configurations to fix transcription errors for standard vocabulary words can also result in mistakes.

Key Facts About Custom Vocabulary Tuning

• Custom vocabulary is only available with CXone Mpower Transcription.
• It works with any language that Transcription supports. If you transcribe in more than one language, all your customizations are saved in the same configuration set. There aren’t separate sets for each language.

• The configuration of a custom vocabulary takes places in the Analytics application in CXone Mpower. You don't need a license for Analytics to customize your Transcription vocabulary, but you do need a license for Continuous Stream Transcription.

• Make sure the characters you use in your custom vocabulary match the language of the transcript output. For example, if your English-speaking contacts often say a Japanese word, you should write that word using English characters in your customization. Using Japanese characters in an English transcript may cause the system to switch to the wrong language. However, if you also support Japanese transcription, you can use Japanese characters in vocabulary customizations meant for Japanese output.