|
Help page
|
|---|
NLP Portal: History and Purpose
latest events first...
| October 2004 |
After five months of implementation, the new portal is ready and
re-launched into a public beta test phase. It has a similar but extended
functionality like the old portal but there is a mighty machinery in the
background:
- The portal uses our PetaMem LanguageServer (PMLS) as backend for the pure
NLP services and "just" handles the duties of a webfrontend.
- Localization is now completedly generic. The portal can be localized to
any language or dialect and to any look&feel.
- There is a complete infrastructure for accounting and pay-per-use
services (including online and offline payment methods). Of course, there
is also a free-contingent infrastructure to enable the casual user or
guest to use a restricted set of this portal for free.
- The portal is running on additional hardware, and has a very elaborated
distributed client-server concept. Main paradigm is scalability of the
system. There is load-balancing and globally distributed servers for
content delivery.
- The portal is prepared to incorporate seamlessly services of partners of
PetaMem, thus further extending and enhancing its capabilities. Some
cooperative projects with partners have started already.
|
|
| May 2004 |
The breakpoint. By this time, a complete specification was made of what
we call the new NLP portal. It was clear, that we needed a portal that:
- would serve as platform for free NLP/NLU services for internet users
- could demonstrate the competence of our company in the NLP/NLU field and
generally serve as demonstration platform for existing or potential
customers
- could be re-used to serve as web-frontend client for our enterprise-class
language processing systems when deployed in large intranets
- would enable us to reach also consumers and generate a complementary
revenue stream on a pay-per-use basis as opposed to selling high-priced
NLP systems.
Some generic code to support the framework and infrastructure needed was also
in place, but it was obvious that the technology used could not support all
of this and a complete rewrite of the whole system was out of discussion.
The main decision: A system based on Mason
technology. Mason is a templating system - that perfectly suited our
needs. Easy integration because of the used programming language, a powerfull
programming concept enabling both re-use of code and fast achievements with
relatively little effort. Moreover it is evidently highly scalable as large
sites like Amazon, AvantGo and others are also based on this technology.
|
|
| October 2003 |
Despite the deficiencies of the old platform, several further
enhancements and bugfixes were made. User management was streamlined and
interconnected with our CRM. |
|
| August 2003 |
By this time, the portal was localized into seven languages, extended
with a Database backend and enhanced with some NLP functionality
(dictionary, language identification, diacritization, numeral
conversion). Two things became clear. First, there were requirements from
our customers of a full online-demonstration of the NLP capabilities of
our language processing systems that were not met. Second, the design of
Yawps was not made for large enterprise-class websites - and exactly that
was where the portal was heading... |
|
| March 2003 |
A first version of the portal - based on the Yawps project - is released. The
main reason for using Yawps were its Perl implementation and a claimed mod_perl (even mod_perl2)
ability. |
|
| January 2003 |
There is an idea to establish "a webpage with dynamic content", that
should provide some NLP functionality and serve an important purpose: The
demonstration of technological competence of PetaMem in the NLP/NLU
domain. |
|
|
