Visualization tool

I really like this :), it's like a more flexible alternative to the music roamer which is one of my favorite stat utilities, I like it because it gives you more control on the graph, you can move any item around freely and you can add any artist individually.

At first I was kind of frustrated because it didn't organized the artists according to their similarity, until I found the cluster button, I think that you should make that clearer it wasn't that obvious, maybe a message somewhere telling you to click 'Cluster' if you want to automatically organize the graph.

I think that It would be cool if you can enter your user name and it create your top 100 or 50 cluster graph automatically similar to this idea

BTW.
How are you calculating the distance between the artists, using the tags or the similar artist list?

I'm going to add it to the overview now

I hadn't noticed that you could add your top 50 artists by adding your self to the graph and then adding all the favorite artists, it was exactly what I wanted :D (top 300 or 100 would be cool too :)), but I have a problem, if I enter my username then show my top 10 artists and finally click 'cluster' the program freezes and doesn't do anything.

Thanks for adding it, the webpage link is probably better than the applet directly, because the people will read the readme and understand more quickly what they can do.

I did what you did and actually this is not a bug issue, just that I have not explained correctly the cluster button.

The cluster button organizes the type of entities that is selected in the drop down box. So, if you add your name, the drop down is user , and than add your artists, and the drop down is still user, so when you press cluster there is only one user to organize, and nothing changes, just the cursor remains changed, but this is a small harmless bug , the program still functions.

You need to change the drop down to artists, and then the artists will be the type of entity that is clustered. For now, one may cluster either tags, users or artists, in the future I may give the option to organize all three categories simultaneously.

Do you think that now the cluster button makes sense, or what changes can you suggest ? Thank you for saying what bothered you, I want to make the program as intuitive as possible.

Have Funn
Gotmagog

Good that I posted in the stats group, the popularity of my appler grows and now I have 500+ hits on my web page for the last week.

I think that my program may get even more popular if people learn to play with it. For example, i read the discussion about tag clouds and posted how they can be generated with my applet

tag clouds with fmcrawler

I have added a new world map location mode, displaying a map and artist/user locations on it. It should be easy and intuitive to use, one just presses the "world map" button to go in map mode and then can press any layout, cluster or voronoi button to get out of map mode.

I would appreciate some feedback on this feature, as well as suggestions for artist'S locations parsed badly.

Have Funn
Gotmagog

-the tooltip bug fixed
-the misfits are OK now
-the advantage fixed
-thursday fixed
-special for USA fans: now locations are shown at exact spot in the USA

such thing will be cool to make for UK as well, which also has quite many music items, but is too small a place

some problems are not so easy to be fixes:
-blindside has many different bands with the same name
--atreyu is set to israel, because they have jericho in their album names, and not a single other word a a clue where they are from

it proved to be more challenging then I thought to make a perfect parser, but it will get better with time.

At least, when one visualizes his friends and other users, it works well, since the location is set explicitly

I read the artists biographies from lastfm, and then try to parse the correct locations from the text. I use a list of names of countries and cities, generated by the Wordnet ontology, and then the first match inside the biography is returned. This may lead to problems if the name of the artist is a country name, or the name of some album occuring before the real country.

I will try to solve this with time and feedback, but this is a challenging data mining task in itself :-). If lastfm makes an artist country webfeed this will sort itself out.

You may open the java console to see some debug information about what string made the program decide for this country, but this would be really boring for anyone besides the program creator :-)

I implemented this right now, seems to be working well.
Thanks for your feedback, it really helps me improve on my program. I alo thought abot wikipedia at first, but I did not expect they have such good information and decided to use other sources. Thanks for reminding me about wikipedia.

If you play some more with the applet and tell me where it makes mistakes now, it will be great.

Have Funn, and thank you one more time for your suggestions.

1. Open:
http://lastfm.atelier-grenzwert.de/

2. In 'Type Press and Enter' select 'User', delete 'type a name ' text, type your username in the box and press enter.

3. Select your picture with left click or drawing a rectangle around it.

4. Right click your picture and select 'favorite artist' from the context menu. Do this step 4 more times in order to add your full top 50 to the graph.

5. Right click your picture and select 'delete'

6. Click 'world map' button in the lower right corner of the page and wait.

7. To navigate the map select 'Transforming' in the lower left corner of the page, and navigate the graph dragging it with the mouse and using the mouse scroll.

Please post here if an artist isn't correctly placed.

I have improved the code some more, better disambiguation of state names in USA and other bugs

i have problem with special unicode characters, like Québec, need to learn how to use them in java

Vulgaires Machins has french wiki, so i won't parse them correctly


the bowling for soup is ok now,others also

OFFLINE