User Test!

[cmouse]

Hello all - we've got a new user test. It would be awesome if some of you took a moment to do it. I honestly need results as soon as possible (measured in hours, not days).

User test

Thanks!

Comments

[anonymous]

I believe its based on GWAP's verbosity. with only difference being that instead of a person on other side you have an algorithm.
I would suggest similar to verbosity where the concept should be given by the interface rather than asking a person to guess.
This way omcs can collect more information about concepts currently under represented in the open mind.
Also this will allow the interface to "Cheat" i.e say if interface is unable to get any questions correct then after few question interface can use analogy space to find most likely question.
Also while a concept is being discovered only yes answer lead to new knowledge thus one cannot have negative polarity statements.

It would be interesting to read the algorithm behind this technique once paper is published.

I guess that one can use hierarchical clustering of properties using property vectors from AnalogySpace. once the tree is constructed it can be used to ask questions using OMCS's infer language. After a certain cluster is determined, other properties belonging to that cluster could be used to guess the concept. though its easier said than done.

I saw that you just changed the interface.
----
Akshay Bhat

[anonymous]

I believe its based on GWAP's verbosity. with only difference being that instead of a person on other side you have an algorithm.

I would suggest similar to verbosity where the concept should be given by the interface rather than asking a person to guess. This way omcs can collect more information about concepts currently under represented in the open mind.
Also this will allow the interface to "Cheat" i.e say if interface is unable to get any questions correct then after few question interface can use analogy space to find most likely question.

while a concept is being discovered only yes answer leads to new knowledge thus one cannot have negative polarity statements.

It would be interesting to read the algorithm behind this technique once paper is published.
I guess that one can use hierarchical clustering of properties using property vectors from AnalogySpace. once the tree is constructed it can be used to ask questions using OMCS's infer language. After a certain cluster is determined, other properties belonging to that cluster could be used to guess the concept. though its easier said than done.

I saw that you just changed the interface.
---- Akshay Bhat