Developers

Build your own Custom Language Model to convert unique Speech to Text

Share this post:

Modern day speech-to-text services are built using a corpus of general, everyday words and pronunciations. While this system works well for common conversation, it can fall short when it comes to accurately transcribing unique accents, industry specific words, or uncommon dialect. The majority of speech recognition services don’t offer tooling to train the system on how to appropriately transcribe these outliers and users are left with an unsolvable problem. Watson is the solution.

Now, with the new STT Language Model Customization capability, you can train Watson Speech-to-Text (STT) service to learn from your input. Create your own custom models that align more closely with your application’s unique requirements, in order to more accurately convert unique accents, topics, and words to text.

Greater accuracy, even with difficult vocabulary

The new STT Language Model Customization capability gives you the ability to generate more unique machine transcriptions, even when the utterances may include highly specialized, industry-specific terminology and jargon.

Here are a few examples of applications that can benefit from Language Model Customization:

  • Medical dictation application. Easily create a custom model by feeding examples of words and phrases the application users might say, including medical terms, drug names, etc.
  • Call-center transcription application. Use custom models to accurately transcribe company names, product names, and industry-specific words and phrases in customer service interactions
  • Retail conversation applications. Create a custom model to transcribe any product name and turn your retail bot into a more effective sales associate.

How it works

The new STT customization API allows a user to add new words (along with their pronunciations in a “sounds-like” style) to the base vocabulary. In addition, the service can also accept corpus files containing sentences from the user’s domain (that is, sentences that a user of the service might say for a particular application). The customization API will automatically extract any new words and provide pronunciations for the new words (which can be adjusted by the user via the API).

Once you input data, the custom model is trained immediately. After training concludes, the custom model is instantly available for use in the STT service.

The STT customization API is a Beta service and initially available for US English only. Visit the Watson speech-to-text page on the Watson Developer Cloud to find out how you can get started using this service.

 

Get started building

Senior Offering Manager, Watson Speech Services

More Developers stories
February 15, 2019

IBM’s Think 2019 highlighted best of AI customer service applications

IBM's Think 2019 was an industry-leading conference that took place February 12-15 in San Francisco and included 2,000+ technical and business sessions with execs from ExxonMobil, Sprint, Honda, KPM, and others. In addition, more than 800 leaders, 400 developers and 200 distinguished engineers from IBM appeared on stage and in private sessions. Speakers ranged from football great Joe Montana, skateboard pioneer Tony Hawk and astronaut Taylor Richardson.

Continue reading

February 12, 2019

Welcome to Think 2019!

Welcome to Think 2019! We decided to try something new this year and update this blog in real-time, so that you may have a glimpse into the conference experience. In case you have not been following along, Think 2019 is IBM's only conference of the year, taking place February 12-15 in San Francisco.

Continue reading

February 11, 2019

IBM Watson is now available anywhere

The team at IBM is thrilled to announce the next big evolution with Watson Anywhere. Now, Watson is available to reach data stored on any cloud—anywhere, making it more open and accessible than ever before.

Continue reading