audio and video meetings involve, by their nature, multiple speakers. This complicates the task of transcription considerably

This is a continuation of my blog post from last week, part of a series of posts on the application of transcription in unified communications.

In this, last week's and next week's posts, I discuss the state of transcription today. In the final post in the series, I'll address where I believe the market may be going and some key areas of innovation that can help us derive more benefit from recorded audio and video content.

Transcribing meetings is, today anyway, a fairly labor-intensive process. Medical and voicemail transcriptions are both complex in their own rights (think esoteric nomenclature and poor reception and background noises, respectively), yet they also benefit from the advantage of a single speaker; in contrast, audio and video meetings involve, by their nature, multiple speakers. This complicates the task of transcription considerably, as it necessitates the reliable tracking of who said what and handling of situations where people talk over each other.

To be effective, meeting transcriptions must be accurate, and they therefore require (at least today) some human involvement in the transcription.

This is, of course, expensive and time consuming. If meeting transcription was offered as a basic service to employees (as audio conferencing or email often are today), I estimate this capability could easily approach 10-20% of an employee's salary at the standard rates of $1–2 per transcribed minute. These rates are based on "typical" turn-around times of three to five days--a span that could reduce the value of the transcription to begin with. Next day transcriptions typically cost three to four times as much.

The costs are understandable; human operators typically transcribe around 30–40 words per minute. Assuming we speak at 100–150 WPM, it can take three to ten minutes of transcriptionist time for every minute of discussion, depending on the complexity of the dialogue, clarity of the speakers' voices and number of meeting participants and the uniqueness of their voices.

There are, however, some innovative approaches to reducing the costs. For example, Cogi allows users to highlight sections of a conference call in real-time to avoid transcribing uninteresting segments (such as, "We've almost got quorum, folks. Can someone give Bob a call on his cell to see if he's going to join?"). I've also seen some creative early uses of Amazon's Mechanical Turk service to accomplish lower cost, faster turn-around transcriptions.

In my final blog post in this series, I'll explore these approaches in more detail and suggest some other areas where innovation can help lead to lower cost, higher accuracy, faster-turn-around, secure transcriptions.

Given the productivity benefits to come from the availability of transcriptions of recorded audio and video content in the enterprise, it's our responsibility as widgeteers, analysts and consultants to ensure the general availability of these capabilities to our customers at a reasonable total burden of ownership*.

* Total burden of ownership considers not just hard and soft costs but end user complexity, an oft-overlooked expense of adopting new technologies.