|  Gavin86 | Mar 27, 2005 3:53pm | I was just curious if anyone here knows anything about semantical logic or any algorithms to look in to that would attempt to discern the proposition (meaning) of a group of interconnected words. Very similar to this very SU software where you can check out a specific interest such as "learning" and see a recommended related topic of "Discovering".
any help would be appreciated :) |
|
|
Sponsor |  artist | Mar 27, 2005 8:03pm | Do you find 'discovering' from 'learning' by finding out the people interested in those topics or via some other way?
Example, you can take the same 5 people who are interested in learning are also interested in discovering. So learning could be related topic of discovering.. |
|
|
| | JC68HC11DLL | Mar 27, 2005 11:25pm | Would seem reasonable since these words are more or less arbitrary as far as any algorithm is concerned, it's probably more meaningful to "define" a word by associating it with what users the word is relevant to, which I suspect SU does depending on what part of SU we're referring to.
I suppose one could set up a dictionary scheme of some sort, but that seems like a lot of work, not to mention high maintenance. |
|
|
Sponsor | |
|
Sponsor | | pjd | Mar 29, 2005 1:02am | | Amazingly, no-one has mentioned WordNet yet? |
|
|
| | JC68HC11DLL | Mar 29, 2005 5:40am | #5 Neat. Nice of them to make it available for download, as it would be frustrating for folks (re)duplicate the effort building dictionaries for any application that required it. But after testing the online version, it appears to be a variant of a thesaurus. It's always going to have holes (as I've found out - missing words) in it that need to be filled over time as any other dictionary scheme would require -- high maintenance.
#4 Tried that url too, any idea how many users there are? It would be much more effective with many users. Did you preload it with interests found on some of stumbleupon's member 'about' pages? |
|
|
Sponsor | | artist | Mar 29, 2005 8:14am | I had used wordnet for my job, as suggestions for keyword searches.
#4, yes, more users, more effective that would be. I haven't preloaded with SU-Members's interests. It would be very nice. I can do that programmatically if there is any long-term interests in it. |
|
|
| | Gavin86 | Mar 29, 2005 1:06pm | ahh, i see i see. thanks for the input everyone. and yes, i built a very crude word relational/reccomendation program, but it did not exactly perform too well, as i quickly realized the relation between the number of users and the accuracy of the results.
keeping an actual dictionary is really not something i'm interested in putting the time or the effort into maintaining or creating. This was really just a side-feature i figured would add to an existing application i'm building, not the crux of the software. still, it's a rather fascinating topic. |
|
|
|
You need to Sign-up for StumbleUpon to post to this forum
|
Algorithms → Discussion