(183) previous ~ index ~ next

---

To: George Doddington <doddington@email.msn.com>
From: Lynette Hirschman <lynette@mitre.org>
Subject: Re: Any CC metrics?
Date: Tue, 15 Sep 1998 11:20:50 -0400

---

At MITRE, Stanly Boykin and Andy Merlino have done some assessment of
closed-caption transcription error rates (word error) vs. careful human
transcription.

I believe it varies considerably by source but it was on the order
of 10%.

HOWEVER, these error rates are misleading if compared blindly
to speech transcription error rates -- in that the closed captioning
needs to be human-intelligible and REAL TIME, as opposed
to complete. So some of the "errors" are due to "compression"
(sometimes the person can't keep up, or just gets the main point)
and quite a few are typos, LOTS of name mis-spellings, homonym
errors, etc.

If there is interest, I can track down our studies and see if
we can make them available.

Lynette

---

(183) previous ~ index ~ next

Last updated Fri Oct 2 19:04:21 1998