the INQLE "Question and Answer Engine"

[David]

More longer term thoughts I have been kicking around. I want to save
these for my later reference.

I had a new idea for the "wiki-like" display of data in an INQLE
server. Semantic wikis have been done. Rather than a wiki, I am more
interested in a question and answer capability. So we develop the
capability to render questions and answers about a given topic. A
semantic question object references the property that it refers to,
like this example for the semantic property:
hasPopulation
We might add a semantic object comprised of these statements:
my:Question1 refersToProperty hasPopulation
my:Question1 hasQuestion "What is the population of
~" (localization:English)
my:Question1 hasQuestion "Was is die Bevölkerung von
~" (localization:German)

This capability for asking questions and rendering answers has several
powerful use cases:
(1) My earlier idea for a survey engine could be powered by this
capability. See http://groups.google.com/group/inqle-core/browse_thread/thread/435d168d6f4d7af7
Through this survey capability, INQLE servers could gather data by
themselves, rather than requiring the administrator to load pre-
generated data.

(2) a new kind of wiki. more Q&A based rather than factual based

(3) a question-based learning algorithm. You tell INQLE the semantic
question of interest and it pursues the answer through gathering data
and data mining

Now INQLE could add intelligence to the process of collecting data, by
determining what it *doesn't* know about a topic. Simpler intelligent
questioning: INQLE thinks: "I know that some countries have the
property "hasGDP", however I currently have no value for the GDP of
Tanzania, so next time the subject of interest is Tanzania, I will ask
about its GDP.

More complex intelligent questioning: INQLE could gather information
much like an expert in some field, asking questions based on its
training on the area of interest. E.g. you could train INQLE to ask
questions to pursue the medial diagnosis of cough. Not simply rules
based but also using data mining and perhaps probabilistic semantic
reasoning. I am imagining that it has access to an expert who can
train INQLE (or perhaps we could use text mining on some interview
transcripts).