Please add the following comment from the
Corpus Analysis list to the message from the Ontology Forum still further below
and draw the obvious conclusions:
To evaluate a system for annotating dialogue acts, you could take a
Discourse Corpus where the text has been manually annotated with
discourse connectives and relations, apply your system to the same
text,
and compare your system's analysis with the manual analysis. For
example,
if you work with Arabic, you could try the Arabic Discourse Corpus
http://www.arabicdiscourse.net/
built by Leeds PhD student Amal Alsaif.
One problem you may find is that analyses are only directly
comparable
if you use the same tag-set of discourse connectives and relations
- eg for the Arabic Discourse Corpus, see
http://www.arabicdiscourse.net/connectives/
and
http://www.arabicdiscourse.net/annotation-scheme/
I assume you want to evaluate accuracy; you might also want to
evaluate ease-of-use, portability, speed etc of your annotation
tool,
and for this you could compare it to other annotation tools eg
Amal Alsaif's READ tool: http://www.arabicdiscourse.net/annotation-tool/
I hope this is useful
Eric Atwell, Leeds
University
http://www.comp.leeds.ac.uk/nlp/
This seems to me to be consistent with the
algorithm below that generates the Self Interest Ontology based on a sample
corpus. Therefore it should help researchers improve the designation of
self interest in texts for various applications or further research on
dialogues.
-Rich
Sincerely,
Rich Cooper
EnglishLogicKernel.com
Rich AT EnglishLogicKernel DOT com
9 4 9 \ 5 2 5 - 5 7 1 2
From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx
[mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Rich Cooper
Sent: Tuesday, May 01, 2012 7:32
PM
To: '[ontolog-forum]
'
Subject: [ontolog-forum]
Self Interest Ontology
The Wall Street Journal has a column
called “the Law Blog”
which could help create ontological theories about self interest based on
observations in law. The present issue is about patents, and the patent
changes passed recently into law.
There are also comments that reflect the
self interest of the participants, which express self interest on various
levels of evaluation, each subjective in construction.
Does anyone have suggestions about how to describe
self interest based on this use case? Any suggestions or comments would
be appreciated. The link is at:
http://blogs.wsj.com/law/2012/05/01/supreme-court-loses-favor-with-the-public/tab/comments/#comment-1214920
The evolution of law, in this example,
patent law, could be traced from the early laws to the present, and the changes
could also be informative about how law changes reflect the interests of the
individuals.
The reason I bring this article up is
described below.
Perhaps a legal ontology could be created
from this information by mining the text from that unifying point of view
– each party chooses behaviors that reflect their individual self
interest(s).
Including self interest among the agents
should help explore and eventually explain the way the data is restructured
with English rules and facts, as well as algebraic and logical
constraints. That also makes it amenable to a discovery structure And/Or
graph.
The discovery structure And/Or graph
provides an ordering of choices made by an ordering of agents. The Solution Forest produced by the And/Or search
algorithm is then available to be validated against a heuristic evaluation
function. Suppose the evaluation function is chosen from among the
Solutions of the And/Or search algorithm.
Validating the evidence would then be a
process of choosing the most useful, or least risky, or most probable, or least
fuzzy, or top ROI choice of solution subtree from the solution forest in the
search.
If I am not making this clear, think of
the evaluation function in an And/Or search which uses a comparison between
alternative solution subtrees stored as a forest of trees within the And/Or
graph.
An evaluation of a filler for a role among
candidate fillers can organize the process by the preferred choice, so that the
choices are ordered (“become primary key indexes” if you prefer
database terminology) and the enumeration of the valuation or arrival order
becomes the primary key for new columns that can describe the various
properties of the enumerated objects. Once the primary key has been
constructed, the arrival order of each choice is already available, and
therefore could provide this key.
The And/Or graph in this example
establishes meaning from the various indexes, i.e., object enumerations.
The Self Interest Ontology generated by this process is represented by the
resulting And/Or graph, the extra columns associated with objects,
relationships, rules and other representations (TBD) are encoded as nodes and
arcs in the And/Or graph and the associated And/Or solution subtrees.
Iterating this algorithmic process will
lead the individual agent to automatically organize a corpus of
documents. Enumerating the iterations leads in turn to a recursion of
evaluation for newly discovered columns to reflect objects, such as choices, or
whatever substructures can be found through the And/Or search forest.
Thanks for any contributions; I am seeking
comments, critiques or suggestions – pick your appellation.
-Rich
Sincerely,
Rich Cooper
EnglishLogicKernel.com
Rich AT EnglishLogicKernel DOT com
9 4 9 \ 5 2 5 - 5 7 1 2