Re: [Fwd: Re: [Fwd: Re: [Phpwiki-talk] Re: RateIt and more feature ideas]]

[Reini U. <ru...@x-...>]

Brad Miller schrieb:
> Dan, Reini:
> 
> I have posted the multilens jar file and cookbook on my website down 
> here at Luther.
> 
> You are free to download the jar and use it under the terms of the GPL. 
>  Over the course of the summer I will add the necessary stuff to the 
> MultiLens source so that we can distribute the source as well as the Jar 
> file (again under the GPL)

Brad,
That's really great news. Thanks a lot!
We will provide a better and personalized MostPopular plugin then :)

http://www.cs.luther.edu/~bmiller

> On Mar 30, 2004, at 1:16 PM, Dan F wrote:
> 
>> Reini,
>>
>> I am bcc-ing GroupLens in case anyone has useful info. As background: 
>> Reini is trying to integrate recommendations into Phpwiki (with my 
>> encouragement!) for 0.5-5 explicit ratings.
>>
>> Apparently Suggest only does binary ratings. Possible options with 
>> suggest:
>>
>> 1. >=3 is 'positively rated', otherwise not.
>> 2. >=user average is 'positively rated', otherwise not.
>>
>> Each of these has the problem that negatively-rated things might show 
>> up in the recs. Then you have to ask for many items and knock out the 
>> negatively rated ones. This is awkward.
>>
>> I will ask around if there are any Suggest versions available that use 
>> rating value.
>>
>> Until then, you might look again at CoFE: 
>> http://eecs.oregonstate.edu/iis/CoFE/index.html. This is a project 
>> headed by Jon Herlocker, a professor at Oregon State who graduated 
>> from GroupLens.
>>
>> I have used neither Suggest nor CoFE, so I can't offer much practical 
>> advice yet, I'm sorry. For MovieLens we use an internal recommender 
>> called MultiLens, by another professor (at Luther College in Iowa) who 
>> graduated from GroupLens, Brad Miller. I have asked Brad for awhile if 
>> he would open-source MultiLens, and he is willing, but has not had the 
>> time to devote to it.
>>
>> Dan
>>
>>
>> -------- Original Message --------
>> Subject:     Re: [Fwd: Re: [Phpwiki-talk] Re: RateIt and more feature 
>> ideas]
>> Date:     Tue, 30 Mar 2004 12:19:43 -0600 (CST)
>> From:     Al M Rashid
>> To:     Dan F <dfr...@cs...>
>> References:     <406...@cs...>
>>
>>
>>
>> Dan,
>>     Suggest only supports binary ratings. And you don't
>>     provide any rating at all, just the item-IDs that
>>     are rated.
>>
>>     -Mamun
>>
>> On Tue, 30 Mar 2004, Dan F wrote:
>>
>>> Mamun,
>>>
>>> Can you tell me the answers to Reini's questions?
>>>
>>> Dan
>>>
>>> -------- Original Message --------
>>> Subject:     Re: [Phpwiki-talk] Re: RateIt and more feature ideas
>>> Date:     Tue, 30 Mar 2004 01:17:25 +0100
>>> From:     Reini Urban <ru...@x-...>
>>> To:     Dan F <dfr...@cs...>
>>> References:     <406...@cs...> 
>>> <406...@x-...>
>>> <406...@cs...> <406...@x-...>
>>> <406...@cs...> <406...@x-...>
>>> <406...@cs...>
>>>
>>>
>>>
>>> Dan F schrieb:
>>> >>> The "Suggest" code is at the following web page:
>>> >>> http://www-users.cs.umn.edu/~karypis/suggest/. It is a C library.
>>>
>>> I'm writing this suggest CGI which gets its input from the mysql data.
>>> However, I don't really grok the suggest library. Hope you can help me,
>>> or please ask george, if not.
>>>
>>> The suggest testdata transactions consists only of binary relations
>>>    user => item
>>> but we have float values per transaction:
>>>    user + item = rating
>>>
>>> gerge writes this in manual.pdf:
>>> userid, itemid
>>> --------------
>>> The actual user-item transactions. These are two integer arrays of size
>>> ntrans such that each (userid[i ],
>>> itemid[i ]) pair represents the transaction that the user stored in
>>> userid[i ] has purchased/accessed the
>>> item stored in itemid[i ]. Note that the users and items must be
>>> numbered consecutively from 0 to
>>> (nusers-1) and from 0 to (nitems -1), respectively. That is, the entries
>>> in the userid[] array can only
>>> take values from 0 to (nusers-1), and the entries in the itemid[] array
>>> can only take values from 0 to
>>> (nitems-1).
>>>
>>> So I ask where or how to pass the ratingvalue to SUGGEST_Init().
>>> Is there no weighted graph supported, only binary?
>>> Or can we pass the values as itemids to SUGGEST_TopN() (converted to
>>> integer)
>>>
>>> Attached is my current code.
>>> Missing: CGI reader. (get the params from env)
>>>           Output

 >Brad Miller, Ph.D.
 >Assistant Professor, Computer Science
 >Luther College
 > http://www.cs.luther.edu/~bmiller
-- 
Reini Urban
http://xarch.tu-graz.ac.at/home/rurban/