[Attempto] reaching utility

[Tobias Kuhn]

Hi,

> AceWiki is a brilliant proof of concept and thanks very much for making it and related technology available.
>
> Now I want to make it useful, to me.  I know a fair bit of ACE and of course i have the preditor but still can't enter the sentences I want into AceWiki.  i think that it may be that the form of the sentence required for entry I wouldn't be satisfied with.  So i ask, Will AceWiki's current approach allow entry of all
> sentences that make sense to try and reason with, in your opinion, for general question answering and relational db front-end querying?  If no, is developing
> a way to allow the user to specify sentence-specific transformations to
> DRS possibly a good idea?  If yes, how? and, should AceWiki attempt to learn from
> these (where they don't conflict with its pre-existing interpretation
> rules)?

So what kind of sentences you would like to add to AceWiki but are not 
supported?

I see that it can be hard to formulate complex knowledge in a way 
AceWiki understands it. I think the predictive editor approach is a good 
way to help users in this process, but it can still be hard.

I don't think that it's possible to support all "sentences that make 
sense to try and reason with". This would probably be equally hard as to 
process full natural language.

The aspect of user-driven extension of the language is a very 
interesting one. Actually, I never thought about this. It could be an 
interesting research area. A problem could be how to generalize from 
ACE/DRS pairs to grammar rules.

> For the short term, it will help matters to be able to map a comment to a set of AceWiki
> sentences so i know exactly what knowledge I was attempting to
> represent. so I think i'll work on that (unless there's a better
> suggestion).

I am not sure whether I understand what you mean. What would happen with 
sentences in a comment that do not comply with the ACE grammar?

> why does AceWiki sometimes disallow a proper noun where a noun is
> allowed?  (Tobias, thanks for your answer to the converse question on
> 12/18/09.)

What's the concrete example? In most cases, nouns are allowed at the 
same positions as proper names but require an article like "a" or "every".

> Can ACE represent any legal DRS?

That depends on what you mean by "legal DRS". The DRSs produced by APE 
are nowhere exhaustively defined. So there is no clearly defined "ACE 
DRS language".

In general, it is easy to come up with a DRS that cannot be produced by 
any ACE sentence. First of all, it is obvious that a DRS must contain 
only the predefined predicates "object", "property", etc. to be mappable 
to an ACE text. But just sticking to the given predicate definitions 
does not necessarily lead to a DRS that has a representation in ACE. 
There are further restrictions about predicates that have to occur 
together and about how variables can be shared.

> Is the RACE source code available?

At the moment, no. But Norbert has plans to release the source code of 
RACE, I think. He can give you more information.


Cheers,
Tobias