By now, every reader should be aware of the
difficulties of efficiently handling the vast amount
of e-mail that comes into their organizations. One
company that has made inroads into solving the e-mail
dilemma is EchoMail. Under the guidance of its
president and CEO V. A. Shiva (who built his first
e-mail system back in 1979 and received a Ph.D. from
M.I.T., where he worked on pattern recognition),
EchoMail has been helping companies (and the U.S.
Senate, which has been known to receive two million
e-mails a day) with their e-mail problems for the past
seven years.
EchoMail's EchoMail product line (available in an
ASP or shrink-wrapped model) employs key-word
filtering to extract attitude, issue, request,
customer type and product/service type from every
incoming e-mail. This allows for a better
understanding of what the customer is asking, how best
to fulfill the request and where to route the e-mail
for response.
Components of EchoMail include: EchoMail Customer
Care, which automatically receives, manages, routes,
responds and tracks inbound e-mail originating from a
Web site or an Internet address; EchoMail/TelePhony (TP),
which enables the integration of e-mail and phone,
including automatic call distribution for blending
e-mail and phone communications; EchoMail/Data
Warehouse (DW), a Web-based platform for data
warehousing of e-mail mailing list data along with a
company's customer database; EchoMail/Reports (RP),
which is a Web-based platform for creating reports
from both e-mail and other customer information;
EchoMail/Direct Marketing (DM), which is a direct
marketing platform for managing mailing lists and
executing outbound e-mail campaigns; EchoMail/Response
Management (RM), which manages responses from direct
marketing promotions and automatically manages bounce
backs, unsubscribes and e-commerce transaction
tracking; EchoMail/Lead Management (LM), which was
built specifically for trade shows and events and
allows clients to gather leads from the trade show
floor, edit the information, post the data directly to
the central database and automatically export the data
through the Web and e-mail; EchoMail/Event
Notification (EN), which enables customers to be
notified based on events such as renewals, billing
statements and other time-based events; and EchoMail/
Business Intelligence (BI), which analyzes each e-mail
that is received or sent and automatically categorizes
each e-mail by attitude, issue, request, sender type
and product/service. Once e-mails are categorized,
EchoMail/BI can be used in conjunction with EchoMail/CC,
EchoMail/DW, EchoMail/LM or EchoMail/DM to
automatically calculate and send responses, integrate
e-mail categorization with existing customer data,
automatically route leads and perform intelligent,
targeted outbound mailings.
Look to voice for a possible solution to the
inefficiency of "while I wondered, nearly napping,
suddenly there came a rapping, as if someone gently
tapping, tapping on my PDA." I've been in meetings
with executives tapping away like so many accountants,
barely looking up from their rude, incessant tapping,
even when you are sitting across the table from them,
highlighting the fact that a major drawback to PDAs is
their time-intensive, inefficient interface.
SpeechWorks, Inc. has released its OpenSpeech Product
Suite, which could not only lead to the voice enabling
of PDA interfaces, but add better voice recognition
capabilities to systems that interact with other
wireless devices as well.
SpeechWorks OpenSpeech Recognizer 1.0 is an open,
high-performance speech recognition engine that
complies with the VoiceXML and W3C requirements. The
OpenSpeech Recognizer processes callers' spoken
commands and ascribes meaning to them. Its standard
features include: natural language, that is it
understands callers' complete phrases and sentences;
speaker independence, that is it understands multiple
speakers including those with accents and without the
need for user training; barge-in, which lets callers
interrupt prompts when they know what they want to do
next; grammar management, where in a VoiceXML
environment, grammars reside separately from the
OpenSpeech Recognizer and are loaded into the engine
only when they are needed to recognize speech; FST
technology, which was developed by AT&T and
incorporated into the OpenSpeech Recognizer to
establish a set of rules that the engine follows when
completing a transaction, and has resulted in up to 90
percent memory reductions compared to previous
SpeechWorks versions and provides the ability to
support vocabularies of up to one million words;
endpointing technology, which incorporates complex
algorithms that ignore high-energy noise events that
are not speech and helps the speech application
determine speech starts and stops; and LEARN
technology, which automatically improves accuracy in
deployed systems by adapting to callers' language
patterns.
SpeechWorks is also providing Open-Speech
DialogModules, which are standards-based VoiceXML
objects. The Collect module is designed to collect
information by prompting the caller and "listening"
for an answer; if none is detected, the caller is
re-prompted. The Evaluate module evaluates a response
to determine if there is sufficient information to
formulate a hypothesis as to what the caller said, or
if the answer should be rejected and the caller
re-prompted. During the confirmation phase, performed
by the Confirm/Act module, a hypothesis with a low
confidence score would ask the caller to confirm or
deny the hypothesis, whereas a hypothesis with a high
confidence results in direct action to the next phase
of the speech application.
The author may be contacted at elounsbury@tmcnet.com.
[ Return
To The August 2001 Table Of Contents ]
|