Speaker 1: Speech-to-text technology has been around a long time, but current tools don't work for everyone. IBM Watson is changing that. Most tools focus on transcription of short messages and search terms from high-quality audio. But what about other sources of audio, such as phone calls, meetings, and broadcasts? What if we could do more with speech? Introducing IBM Watson Speech-to-Text. IBM Speech-to-Text uses advanced statistical modeling techniques developed over decades in IBM research and refined by ideas from cognitive computing to transcribe both high-quality and lower-quality audio from a wide variety of source material. It uses the technology behind Watson to automatically determine the most accurate result for words and phrases and present them with confident scores and other metadata. Call centers could transcribe millions of minutes of recorded audio automatically, allowing them to mine the information in the calls to identify issues to provide more value to their customers and agents, no more having to furiously take notes during lectures and meetings. Speech transcription allows you to actively pay attention to the discussion. When the meeting is complete, a transcription could be waiting for you, whether it be from an audio or video recording. Using this technology, the full content of an entire library of recordings could be made searchable without the need for human tagging. When combined with other services on the Watson Developer Cloud, you can build even better cognitive applications.

Speaker 2: Watson Speech-to-Text is an API-based service that is specialized for converting human voice into text featuring a special data format. The data that is returned includes not only the translated text, but also alternative translations along with the confidence scores for each one of those translations. Out of the box, the service can translate general utterances and works with many commonly used phrases. If you need Watson Speech-to-Text to understand less commonly used words and phrases, perhaps those specific to your industry, the service can be trained to recognize domain-specific terms. IBM provides access to a number of software development kits available on GitHub for working with the Speech-to-Text service. Since the service is hosted on the IBM Cloud, it is scalable, allowing you to have multiple services working in concert to translate very large numbers of speech into text. The service is highly customizable. It can be trained through the API to recognize many words and phrases that may be specific to your use case. Watson Speech-to-Text also supports a number of ways to connect to the service with either live streams or pre-recorded audio. Most importantly, all of the data that passes through the Speech-to-Text service is your data.