Ed, (01)
Thanks for bringing that example to our attention. I followed the trail
to the free viewer for Open Calais: (02)
http://viewer.opencalais.com/ (03)
If you go to that site, you can type or paste any text, and the system
will display a version that highlights the "named entities", key terms,
and coreferences. It also identifies the topics and other info. (04)
As an example, I pasted the text below, which I extracted from an
article on the ScienceDaily web site. Following is the full article: (05)
http://www.sciencedaily.com/releases/2007/09/070906140753.htm (06)
If anyone would like to see what Open Calais does with that text,
cut and paste the excerpt below the line after my signature into
the window on the viewer.opencalais.com web site, and press "Submit". (07)
The results are interesting. The program can give some useful
information, but it needs a lot of work. For example, it identified
the topic of that text as "Environment" with a 61% rating or
"Health Medical Pharma" also with a 61% rating. (08)
The ScienceDaily web site had a much better classification of related
topics. It lists the following tree of topics and subtopics, any of
which the reader could click for related info: (09)
> Plants & Animals
>
> * Monkeys
> * Apes
> * Endangered Animals
>
> Fossils & Ruins
>
> * Human Evolution
> * Anthropology
> * Early Humans
>
> Reference
>
> * Mirror neuron
> * Monkey
> * Primate
> * Capuchin monkey (010)
The classification of named entities was not bad. It underlined and
identified organizations, such as "Harvard University" and persons
such as "Justin Wood", "Marc Hauser", and "Brenda Phillips". It also
correctly identified "Wood" as a reference to "Justin Wood". (011)
However, its resolution of pronoun referents was rather poor. For the
following sentence in the text, (012)
This indicates that the primate inferred goal-oriented action on
the part of the experimenter when he grasped the container, and
was able to understand the difference between the goal-oriented
and accidental behavior. (013)
The correct referent of the pronoun 'he' is the noun phrase
'the experimenter', but there is insufficient information in the
text to determine whether 'the experimenter' refers to any of
the named persons in the text. The OpenCalais system, however,
identified Justin Wood as the referent of 'he'. (014)
A later sentence in the text has four occurrences of 'his', and
the Open Calais system incorrectly identified Justin Wood as the
referent of all of them. The last sentence has two occurrences
of 'them' and two occurrences of 'their', but Open Calais did
not attempt to resolve them. (015)
In summary, the system does something useful, but a lot more
work is necessary. I don't know whether ScienceDaily generated
their topic tree automatically or semi-automatically, but it is
certainly much better that the two suggested topics by Open Calais. (016)
John Sowa
_____________________________________________________________________ (017)
When trying to understand someone's intentions, non-human primates
expect others to act rationally by performing the most appropriate
action allowed by the environment, according to a new study by
researchers at Harvard University. (018)
The work was led by Justin Wood, a graduate student in the Department of
Psychology in the Faculty of Arts and Sciences at Harvard, with David
Glynn, a research assistant, and Marc Hauser, professor of psychology at
Harvard, along with Brenda Phillips of Boston University. (019)
"A dominant view has been that non-human primates attend only to what
actions 'look like' when trying to understand what others are thinking,"
says Wood. "In contrast, our research shows that non-human primates
infer others' intentions in a much more sophisticated way. They expect
other individuals to perform the most rational action that they can,
given the environmental obstacles that they face." (020)
The scientists studied the behavioral response of over 120 primates,
including cotton-top tamarins, rhesus macaques and chimpanzees. These
species represent each of the three major groups of primates: New World
monkeys, Old World monkeys and apes. All three species were tested in
the same way, and the results showed the same responses among the
different types. (021)
In the first experiment, the primates were presented with two potential
food containers, and the experimenter either purposefully grasped one of
the containers, or flopped their hand onto one of the containers in an
accidental manner. For all three species, the primates sought the food
container that was purposefully grasped a greater number of times than
the container upon which the hand was flopped. This indicates that the
primate inferred goal-oriented action on the part of the experimenter
when he grasped the container, and was able to understand the difference
between the goal-oriented and accidental behavior. (022)
In the second experiment, the researchers asked if the primates infer
others' goals under the expectation that other individuals will perform
the most rational action allowed by the environmental obstacles. Again,
the primates were presented with two potential food containers. In one
scenario, an experimenter touched a container with his elbow when his
hands were full, and in another scenario, touched a container with his
elbow when his hands were empty. (023)
The primates looked for the food in the container indicated with the
elbow more often when the experimenter's hands were full. The primates
considered, just as a human being would, that if someone's hands are
full then it is rational for them to use their elbow to indicate the
container with food, whereas if their hands are empty it is not rational
for them to use their elbow, because they could have used their
unoccupied hand. (024)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (025)
|