This web app
remains under construction. As a preview: it works like PubMed
Alerting, using a similar Register
page to create similar support objects that can import data items, then monitor and constrain matches.
In a Lexicon service,
the main data objects are definitions for vocabulary items registered in digital
lexicons. They can document
and help publish users' schema,
data dictionaries,
controlled vocabularies,
ontologies,
etc. much like PubMed abstracts can
help publicize research results.
Our web software uses the same mechanics to search over either type of
data.
In both services, a unique symbol (PMID or sense ID) will be summarized
by some properties and some well written
English (an abstract or a definition).
Here, however, a query locates published usage specs for a concept, including its spellings in one or more natural
languages. Clicking on such a
spec in one lexicon
will show similar specs in others. The best matches, if well constrained, should work as
default equivalents for translation
or ontology-aligning web
services.
Sense-finding and PMID-finding both
exploit upper ontologies that help
model meaning for spellings in indexed text. The UMLS' Semantic Network
is biased
toward Life Science topics, and Lexikos can
integrate others - Idealized English
and Roget's Thesaurus - which
it deems useful for more generalized NLP.
Subscribers may add new terms, provided each concept is registered
under upper ontology discriminants.
Wizards make this easier, and your
low level concepts can inherit from
your higher ones, further cutting the work.
Textual definitions must not contradict such formal tags, and must meet
site-wide style standards.
Such documentation practices are consistent with professional
lexicography and curation. We hope our tools will help groups to
integrate domain ontologies,
by organizing and adding needed upper ontologies,
constraints and consistency.
This service will not eliminate interoperation
problems, but it will
guide groups who want to share vocabularies to better spec their own
terms, find semantically
similar ones, expose referential
differences, and get ALERTS about all relevant progress
as formal differences get gradually reduced or accommodated.
Policy wise, semantic web services should stop CRAFTING
machine-readable vocabularies and
instead try to systematically MANUFACTURE them,
using tools and techniques like these, specifically
intended to support their
machine-aided bulk integration and
semi-automated
growth.
Please email me to discuss participating in open-source
lexicon development
that can integrate vocabularies useful in your own
work, and I will walk you through some additional design props not yet
exposed
here, showing how matches can be made to depend on the evolving context of a specific document or
dialog.
|