 |
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
FOR
IMMEDIATE RELEASE UISPEECH
INTRODUCES ENHANCED "MEDIA MINING SYSTEM" TO BROADCAST INDUSTRY AT
THIS YEAR'S NATIONAL ASSOCIATION OF BROADCASTERS CONFERENCE Revolutionary
New Speech Recognition and Archive Retrieval Technology Set To Reduce Editing
Costs and Crucial Man Hours Through Innovative Technological Platform March
11, 2003- Los Angeles, CA—UISpeech, a division of Uptime
Integrated Systems (UIS), will be introducing an enhanced Media Mining
Voice-Recognition Technology System at this year’s National Association of
Broadcasters Conference, it was announced today by Ken Stirbl, President of
Uptime Integrated Systems. The innovative technological platform is set to
transform the way business is conducted within the broadcasting, newsgathering,
motion picture, and cable and satellite industries, among others, by new
cost-cutting and efficient methods of handling media content.
This
technology was previously used within the government intelligence community for
surveillance and intelligence gathering, but commercial applications have been
recently discovered for this technology, including broadcast monitoring,
broadcast news indexing and editing, digital media asset management, media and
financial analysis, interactive news services and much more. The Media Mining
System was created by Sail Labs Technology, and is being launched by UISpeech
within the United States. UISpeech is responsible for integrating
and implementing the Media Mining Technology into new or existing systems,
corporate networks and Oracle Data Centers within the broadcast industry.
Additionally, UISpeech will develop and identify new commercial
applications for this system. “UISpeech and Uptime Integrated Systems have determined
after years of testing and refining that this product is ready to go. We are
confident that this technology will add substantial value to organizations from
diverse industry backgrounds and now is the time to run with it,” said Stirbl.
The technology behind the Media Mining System enables media streams (audio, video, satellite, etc.) to be received through the Media Indexer which then are passed through a series of technologies utilized by the Indexer including speaker change detection, speech recognition, speaker identification, named entity detection and topic detection. Speech recognition technologies are utilized initially, which breaks down transcripts so that the input can be converted to an indexed XML output file for archiving and retrieval as needed. All files are uploaded to the Media Mining Server and are accessible through the Media Mining Explorer, the graphical user interface that allows the user to search, summarize, and replay audio/video content. The
highly advanced speech recognition technology within the system utilizes
a context-sensitive statistical model of natural language to identify the
individual voice characteristics of different speakers. Spoken words are
analyzed and evaluated according to their position within a sentence rather than
relying on the phonetic spelling of a word. The system is equipped with a
vocabulary of between 65,000 and 500,000 words, depending on language selection,
as well as over 400 million probability algorithms of broadcast based text
sentences, which ensures high recognition accuracy. The UISpeech system
has an average accuracy rate of 87.5% for general live broadcast news, but
results have been seen as high as 98% accuracy for specific speeches for
controlled recorded environments, as compared to other systems whose accuracy
rate ranges from 40-50%. These added features can help a user search for and
identify speaker specific sound bytes that are easily accessible through a
user-friendly graphical user interface. UISpeech has also developed a new
learning tool that is able to increase the recognition accuracy for specific
vertical market segments. This
platform provides significant benefits to the broadcasting industry, including
indexing and archiving live broadcasts in real time and recorded media in real
time or better, closed captioning, automatic keyword translation, advanced
searching capabilities and creating edit decision lists (EDL’s). The timing of
information retrieval is crucial for breaking news stories, and the Media Mining
Technology expedites this process by enabling the end-user to search for
pertinent audio/video content in multimedia sources in real time ARCHIVING AND INDEXING Broadcast
news organizations handle immense volumes of information on a daily basis.
Indexing and annotating this information manually consumes several productive
hours of manpower. The Media Mining System is a highly sophisticated system that
automates this process. Information from multiple sources can be indexed and
archived in real time or better 24 hours a day, 7 days a week resulting in
higher productivity and cost savings for the broadcasting company. This
archiving tool manages a company’s media in a more effective way. Closed
Captioning At
present, the closed captioning business is constrained by a few factors. Manual
closed captioning of news broadcasts, television series or movies, requires
considerable time to complete the transcribing of scripts, timing and placing of
captions, and encoding processes. Human error can also account for inaccuracies
in captioning, as well as relying on pre-scripted teleprompter coverage. The
closed captioning solution offered by the Media Mining System greatly reduces
the need for manual processing. The UISpeech system automates this
process, which annotates and archives information continuously while
simultaneously displaying closed captioning. Automating this procedure allows
more information to be processed quickly and accurately and ensures
that no spoken data is lost within newscasts, sporting events, and other
breaking stories due to that spoken text not being included in the
original script. KEYWORD
TRANSLATION An
integral tool to monitoring international broadcasts is keyword translation. The
Media Mining System allows users to create English language queries for Arabic,
Chinese, French, German, and Spanish and vice versa. These queries search for
specific words within the foreign language text and alerts a user to the
appearance of that word within the foreign text according to a color key. No
prior knowledge is needed of these foreign languages as a user can search for
translations of specific keywords within the broadcast. For security purposes,
this information can be tracked and logged. Advanced Searches
Information
storage is structured according to predefined topics and subcategories within
those topic areas, which are displayed in an archive tree format. The advanced
search function allows a user to search by person, place, topic, location,
organization, date, time, percentage and text. Narrowing the focus of a search
to channel, time, or language can further refine searches for the broadcast
monitoring industry. Based on the search criterion selected by the user, the
search engine will search through folders within the archive and return a list
of stories and associated media files for processing the advanced queries. EDIT
DECISION MODULE UISpeech
was directly responsible for developing the Edit Decision Module, a major front
end of this system that has proven beneficial to the editing, production and
post-productions processes. Formerly, searches were conducted manually for sound
bytes to find spoken dialogs, transcripts had to be read, time codes had to be
researched, words and video had to be separated and logged manually into the
existing editing tools. Utilizing the Edit Decision Module, an Edit Decision List (EDL) can be created, which
allows a user to cut and paste audio/video content with time codes for
live-to-air playback capabilities. This has proven to be highly valuable for
producing movie trailers, spots and news pieces. Story creation can be
facilitated by virtually anyone from anywhere using this application.
About UISpeech
UISpeech, a division of Uptime Integrated Systems, Inc (UIS), is at the forefront of speech communications technology for the broadcasting industry, in addition to other diverse industry sectors. UISpeech is able to customize and integrate Sail Labs technology with other companies’ technologies to satisfy the unique requirements of each client in media mining. Our primary function is to integrate Media Mining and Speech Recognition Technology into new or existing systems, corporate networks and Oracle Data Centers. UISpeech will create custom Interfaces and Enhanced support function modules, which will get the most out of the Media Mining Technology. For more information visit www.uispeech.com. |