News 6th Sep 2018

Feature launched to differentiate between how a word sounds and how it is written

Speechmatics has launched the Sounds feature, a new addition to the company’s current speech-to-text offering. Sounds supports broadcasters by delivering highly accurate transcripts, it allows the speech engine to understand the difference between word pronunciations and the way they are written.

Sounds extends the Custom Dictionary feature, which lets users add context-specific words in real-time, such as footballer names or breaking news locations, instantly delivering consistency and reliability of specific words and ultimately increasing the accuracy. By refining the pronunciation within Custom Dictionary, Sounds can help with the spelling of names, products, acronyms, abbreviations, trademarks, copyrights and alternate word forms.

Ian Firth, VP Products at Speechmatics, explained: “In the broadcast industry, subtitling for names and words that don’t sound phonetically as they are written is an ongoing bugbear and can be a cause for significant embarrassment by the broadcaster. For example, knowing how to spell Condoleezza Rice’s name flawlessly every time. With Sounds, our engine simply needs pronunciation hints that describe the sounds of the words, like ‘AI’ pronounced as ‘ay eye’. It is not necessary to use phonetic definitions, but something that sounds like the word you want to define and the way that you want it to be written. It can even be used to change things completely, for example you could easily configure it so that ‘Dr’ is written every time ‘Doctor’ is said.”

With other vendors’ offerings, a pronunciation pack is required when using a hints-style feature. With the Speechmatics solution, Sounds is built to enhance the speech engine when required and supporting multiple pronunciations of the same word. As the solution can be delivered in private environments, it is highly secure and accessible only to the broadcaster.

Firth continues, “Accuracy is still the key metric for speech recognition systems and remains top of mind throughout the development of our solutions at Speechmatics. At Speechmatics, we pride ourselves on unparalleled accuracy rates, and Sounds enables us to continue to improve on our speech-to-text accuracy by building technology that understands pronunciations that are personalised to the user and use case when needed.”

The Sounds feature is available now through Custom Dictionary. To try a demo of the Sounds feature at IBC in Amsterdam 14-18 September, visit Speechmatics at stand 8E17.

X