[Attempto] Pellet vs RACE

[George Herson]

Hi Tobias,

For layer 2, if complex statements not likely to involve gobs of memory could be 
identified, i'd start them right away at low priority rather than waiting, to 
conserve cpu and user time.

As you say, retaining performance with scale has to be solved to achieve non-R&D 
utility.  What about the idea to stick the wiki data in a SQL engine?  Is 
modern, Turing-complete SQL too incapable somehow compared with Prolog?  Would 
Datalog be a cost-effective compromise?

On the front to expand AceWiki's ACE subset, is RACE ready for release Prof 
Fuchs?  I volunteer to work on giving it the needed Java API.  The stateless 
RACE Web Service isn't a good alternative for AceWiki given the amount of text 
that would need to be sent with each request.

thanks,
George

----- Original Message ----
From: Tobias Kuhn <kuhntobias at gmail.com>
To: George Herson <gherson at snet.net>
Cc: attempto at lists.ifi.uzh.ch
Sent: Tue, January 19, 2010 10:41:34 AM
Subject: Re: [Attempto] Pellet vs RACE

Hi,

I agree that the bad performance of AceWiki for complex or relatively 
large knowledge bases is a big problem. This should be solved in one way 
or another...

Heuristics that retain correctness and completeness are anyway applied 
by Pellet. I am not an expert in reasoner implementation and I don't 
know all details of Pellet, but I think that it would be a big effort to 
improve this, since a lot of work has gone into Pellet already to 
improve its performance.

Heuristics that do *not* retain correctness and/or completeness could 
really improve performance, I guess. But I am not sure whether this is 
the way to go.

Another approach I was thinking of is to have a layered ontology:
- Layer 1: Very simple statements only. Reasoner runs immediately.
- Layer 2: More complex statements. Reasoner runs every 24h, or so, in a 
parallel process.
- Layer 3 (red statements): Very complex statements. No reasoning within 
the wiki.

There could be even more layers. It's just a vague idea so far. I am not 
sure whether it would work out...

Anyway, it's definitely a crucial issue if systems like AceWiki should 
be practically used somewhen in the future. Any ideas are appreciated.


Cheers,
Tobias


On 01/18/2010 05:19 PM, George Herson wrote:
> Tobias,
>
> Regarding AceWiki slowing down when the subset of ACE supported increases, 
>there must be cost-effective heuristics we can use, like sectioning the data to 
>quickly find the subset most likely useful with a given problem?  (Others?)  
>Also, we can use MySQL etc to store large fact bases 
>(http://tinyurl.com/SQLfasterthanProlog).  Do you agree?
>
> thanks,
> George
>
> ----- Original Message ----
> From: Tobias Kuhn<t.kuhn at gmx.ch>
> To: George Herson<gherson at snet.net>
> Cc: attempto at lists.ifi.uzh.ch
> Sent: Mon, November 30, 2009 6:33:56 AM
> Subject: Re: [Attempto] Pellet vs RACE
>
> Hi George,
>
> There are several reasons for the fact that AceWiki uses Pellet instead of RACE 
>(but we have some plans to enable the use of RACE within AceWiki in the future).
>
> First of all, RACE does not have a Java interface at the moment, which makes it 
>harder to integrate it into AceWiki.
>
> Furthermore, performance would (potentially) suffer when using a larger subset 
>of ACE (as the one supported by RACE) than the OWL subset (as supported by 
>Pellet). I agree that it would be desirable that AceWiki can do reasoning within 
>a larger subset of ACE than the one supported at the moment. But already now, 
>AceWiki can get quite slow when the knowledge base gets large. And switching to 
>RACE would enlarge this problem.
>
> Another problem is that RACE does not provide (at the moment) an API for 
>retrieving the set of instances for a given class, or for extracting the class 
>hierarchy.
>
> Long answer short: AceWiki could in principle use RACE but it uses Pellet at 
>the moment because of performance and interface reasons.
>
>
> Cheers,
> Tobias
>
>
> George Herson wrote:
>> Hi all,
>>
>> Why does AceWiki use the OWL-based Pellet reasoner instead of RACE?    OWL's 
>>expressiveness is a subset of ACE's.  The more of AceWiki's content that can be 
>>used by the reasoner the better, i think.
>>
>> "The OWL reasoner Pellet<http://clarkparsia.com/pellet>  is integrated into 
>>AceWiki. It ensures that the content of the wiki is always consistent. The 
>>reasoner is also used to infer the class-hierarchies and class-memberships, and 
>>the results are displayed in ACE again. Furthermore, questions can be added to 
>>the wiki articles. Those questions are evaluated by the reasoner [...]" -- 
>>http://attempto.ifi.uzh.ch/acewiki/
>>
>> thank you,
>> George
>>
>>
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> attempto mailing list
>> attempto at lists.ifi.uzh.ch
>> https://lists.ifi.uzh.ch/listinfo/attempto
>