InfraWare Speech Recognition Technology Improves Productivity

InfraWare recently announced the 3rd generation of our back-end speech automatic recognition service (ASR) technology.

InfraWare’s First Draft is a back-end automatic speech recognition (ASR) service geared specifically toward transcription of medical dictation. The First Draft service doesn’t replace medical transcriptionists (MTs) but rather provides them with a power tool to produce more. It is one of InfraWare’s many technology solutions designed to pull time and cost out of the transcription process.

The service runs in InfraWare’s datacenter under the application service provider (ASP) model of software licensing. And, it is much more comprehensive than traditional speech recognition. The service begins with great ASR and builds from there. Since the process actually happens in a data center server where it isn’t visible, the results are first observed in the InfraWare Transcription Client (ITC), where transcription and editing are performed.

Building on InfraWare’s legacy of back-end speech recognition, the new First Draft service is more accurate than ever ‐ and it now recognizes more than just speech and basic context. The business value comes from bridging that gap between good ASR and the requirements of quality transcription. For example, most physicians don’t dictate periods. Even if word recognition were perfect, that leaves a lot of cognitive, time-consuming work for an MT to find sentence boundaries. Then, a fresh reading is required to find internal punctuation like commas.

The artificial intelligence (AI) behind the new First Draft can find many of those periods, appropriately correct the capitalization and even insert some commas. Text is placed in the document template under the appropriate heading and synchronized with the audio playback for easier editing. The result is still an imperfect document, but significant time is shaved from the production process by making the draft easier to read and understand. The AI engine developed to add this value provides similar recognition for numbered lists and other formatting elements common in medical reports but not dictated.

Most speech recognition products learn over time. In particular, they learn how an author sounds when they dictate. The new First Draft service takes that learning to an entirely new level by learning what authors dictate and when. By accumulating and applying this knowledge, the improvements that come from learning are deeper and more effective. And, unlike some speech recognition vendors in the transcription industry who have given up the WYSIWYG (what you see is what you get) editor environment, InfraWare has accomplished all these advancements while staying true to our roots in a powerful transcription-centric editor that displays like it prints.

What is back-end speech recognition?
Back-end recognition describes technology that allows physicians to dictate without changing their behavior (without interacting with recognition software) and performs ASR in the datacenter sometime after dictation and prior to transcription for the purpose of accelerating the transcription process. InfraWare first introduced back-end speech recognition in 2004. Over the years, the company has improved the performance of the First Draft service through three generations of innovation.