Category : Various Text files
Archive   : NLM-INFO.ZIP

Output of file : AUDIOKNO.TXT contained in archive : NLM-INFO.ZIP
National Library of Medicine

The Audio Knowledge Acquisition Tool

The National Library of Medicine (NLM) has developed a hardware and
software system for the management of complex verbal protocol data. The
system, called the Audio Knowledge Acquisition Tool (AKAT), is a
Macintosh-based tool for people who use audio recordings in their daily
work. It offers an alternative to transcribing those recordings by
allowing users to describe and organize the content of the recordings
on-line, with the capability of finding and playing any desired audio
segment instantly. There are many different potential uses for the
AKAT. It was originally developed to help knowledge engineers analyze
recordings of experts thinking out loud as they solved problems.
Psychologists who do protocol analysis and psychiatrists working with
clinical recordings will also find it useful. With AKAT, audio
recordings can be as easy to manipulate and analyze as are their

AKAT can be more than a protocol analysis tool. Oral historians, sound
archivists and others who are curators or analysts for large collections
of audio information will also find it useful. It can replace or
supplement expensive, time-consuming transcriptions, which almost
inevitably lose important information found in tone of voice,
paralinguistic utterances and phrasing. It has potential educational
applications as well. Students can match their course notes with
recordings of the lectures that they've attended. They can reorganize
their notes in whatever way makes the most sense to them, and always be
able to find any part of the recorded lecture instantly. Teachers can
associate detailed notes with their lectures, and even add Macintosh
graphics as annotations.


Basic research into machine learning, expert systems and artificial
intelligence, generally has relied on the analysis of protocols of human
problem solvers thinking aloud as they do their work. These protocols
are then transcribed, organized, and analyzed to identify the concepts,
relationships and processes involved in the expert's performance. The
cognitive elements identified in this manner are combined to form a
model which is used as a basis for the design of automated analogues.
Protocol analysis also plays a significant role in other areas of
biomedical research including psychiatry, neurology and the history of
medicine. The AKAT grew out of the protocol analysis needs of the
Machine Learning Project at the NLM.

How AKAT Works

AKAT is capable of digitizing audio from tape recordings or other
sources, storing the digitized audio on magnetic or optical disk, and
providing a means of associating typed annotations with points or
segments in the audio. Annotations can be freely moved and can be
organized into hierarchies and groups. The audio associated with any
annotation can be instantly recalled and played. Annotated audio files
can thus be rapidly browsed for relevant information without losing
immediate access to the underlying audio.

The AKAT is an application program that runs on Apple Macintosh_
computers, using inexpensive, commercially available audio digitizers.
It has an intuitive, Macintosh compliant user interface which makes it
very easy to learn and powerful in use. The AKAT requires a Macintosh
II series or Macintosh SE computer with at least 1 megabyte of memory,
running System 6.0.3 or higher. Currently, the sys-tem also requires
either a Farallon MacRecorder_ or a Digidesign Audiomedia_ audio
digitizing device. Due to the large size of digital audio files even
if compressed, a removable media disk drive, WORM optical disk drive or
magneto-optical disk drive is strongly recommended. A 45-megabyte
removable media disk holds about 90 minutes of audio and annotations.

The Advantages of AKAT

Typically, verbal protocols are transcribed before they are analyzed.
The transcription process is expensive and time consuming. On a complex
project there may be hundreds of hours of protocols gathered. The
advantages to be gained by integrating the auditory material directly
with the analytical tools can be measured both in terms of cost-
effectiveness and increased quality of the knowledge acquisition.

First, because the protocols are generally extensive and contain complex
and specialized language, transcription can easily cost thousands of
dollars per project. Specialized transcription personnel must often be
utilized. Because the additional equipment and software needed to
eliminate the transcription step is relatively inexpensive, significant
savings can be realized even within a single project. Amortizing the
cost of the equipment over several projects makes even greater savings

Second, valuable information available in the purely auditory portion
of the protocol is made accessible to the knowledge engineers. This
information is often necessary to make sense of the words in these
protocols. The expert systems development literature suggests that this
nonverbal information is significant, but is rarely used because of the
difficulty of managing it. Improving the quality of the knowledge
acquisition process with this audio protocol analysis tool may speed the
construction and improve the quality of resulting systems.

Finally, a system for managing and analyzing audio materials has
significant potential for additional developments and other uses. In
particular, systems for annotation and analysis of video images will
become more important as the cost of acquiring and storing such images
continues to fall. Although it is not yet cost effective to develop a
similar system for managing videotaped protocols, the development of an
audio system should pave the way for a follow-on video system. Research
in compression technology at the Lister Hill National Center for
Biomedical Communications and rapidly developing technologies for use
of real-time video on the Macintosh suggest that a video knowledge
acquisition tool may be plausible in the near future. The interfaces
and many of the other components of audio and video systems would be
closely related.

Project Status

The significance of the audio knowledge acquisition tool lies not only
in its application to the elicitation and development of cognitive
models for machine learning, but also in knowledge engineering for
expert systems, in the organization and analysis of oral his-tory, in
education and in other fields where rapid, organized access to large
amounts of audio information is important. It is our goal to transfer
this potentially valuable technology developed at NLM as broadly as

The AKAT is now being tested at a variety of sites throughout the United
States, including in-house at NLM, at a major medical artificial
intelligence research center, at a university collection of oral
history, at a national museum and in an innovative secondary school
teaching program. Ongoing feedback from these test sites has been
extremely helpful in improving the tool. Critique by a well known
computer graphics design firm has contributed to the design of the AKAT
user interface.

Pre-release improvements are still being made in the interface, the
audio data compression routines and various other aspects of the system.
Additional modifications to the program are also under way, based on
expected improvements in the multimedia capabilities of the Macintosh
operating system and in changes in available audio digitizing equipment.
The possibility of porting the tool to other computer platforms is also
under consideration. The NLM hopes to make a version of the AKAT
available to the general public during calendar 1991.

For further information, contact:

Audio Knowledge Acquisition
Tool Development
Computer Science Branch
National Library of Medicine/LHNCBC
Bethesda, Maryland 20894

  3 Responses to “Category : Various Text files
Archive   : NLM-INFO.ZIP

  1. Very nice! Thank you for this wonderful archive. I wonder why I found it only now. Long live the BBS file archives!

  2. This is so awesome! 😀 I’d be cool if you could download an entire archive of this at once, though.

  3. But one thing that puzzles me is the “mtswslnkmcjklsdlsbdmMICROSOFT” string. There is an article about it here. It is definitely worth a read: