As was foretold, we've added advertisements to the forums! If you have questions, or if you encounter any bugs, please visit this thread: https://forums.penny-arcade.com/discussion/240191/forum-advertisement-faq-and-reports-thread/
Options
Word clouds
Antoine
__BANNED USERS regular
__BANNED USERS regular
When you mouse over a thread title a little word cloud should pop up (similar to how the first post's text pops up) to help you see what the hell the thread is really talking about at the moment.
Also makes it easy to see if a thread has derailed hard, without really reading the posts.
There's just too many threads to read these days.
Also makes it easy to see if a thread has derailed hard, without really reading the posts.
There's just too many threads to read these days.
Antoine on
0
This discussion has been closed.
Posts
This is what you would see, as you moused over a thread.
Just by looking at those words, and their popularity indicated by their relative sizes, could you guess what the thread is talking about?
Yup.
But would you have to read a single post to figure that out?
Nope.
What does this mean for PA?
It means mods can better identify the threads that are going way too off topic. For instance, that Olbermann thread is probably going off topic. There's some shit in there that doesn't make sense. This saves time, and thus, saves money.
It also helps us, the posters. Many times, we don't even read posts. We just skim them, and we skim them for a while, looking for threads that look interesting. Wouldn't it be great if we can just "taste" a thread before we even open it? It would allow us to find good threads quicker, without reading a bunch of bullshit first.
This is especially useful too for stuff like chat threads or pretty much any thread in SE++, where the topic may be very volatile and subject to change from post to post. A word cloud would give us an indication of where the thread is at.
Ha, that's incredibly naive.
We are in the year of two thousand ten, in the third millennium of human history. Threads and conversations are no longer moving at mere internet speeds, they are moving at mobile speeds.
Thread posts are like stars, the earliest posts in the thread are like the light of distant stars just reaching us. They look up to date, but in fact, those thoughts and conversations died long ago, and we are now talking about new things.
1) you'd have to iterate through every word in the thread - this is just as intensive as running a search
2) and just like search, you'd have to drop indexing 2 letter and 3 letter words
3) to keep it from killing the forum, you'd have to cache the 'tag cloud', and in fast moving threads this would make the Tag Cloud of Relevance lag too far behind to be useful.
We DID have a manual tagging feature that came as part of the new forum upgrade, but it was disabled. Maybe it can come back some day? (It probably won't, though)
Not really that intense.
See what you do, is everyone time someone makes a new post or edits a post, you iterate through every word in their post, and use the word as a "key" for a hash table of values. Basically everytime a word is found in the hash table dictionary, you see if it has a stored count value, and if it does, increment by 1, and if not, set the value to 1. Or something.
Later, you would take the values and display the word cloud words based on some parameters, for instance, words would start showing up after a certain count value (which may or may not depend on how long the thread is), and the font sizes of the words would increase as the value gets higher.
The PA forums are a really big and really intensive thing that have to live alongside and not impact the front money-making page of the site at all in any way shape or form.
There's a reason why the search timer is almost two minutes long (well, the last time I actually used in-forum search anyway - a lot of people just use google search with parameters set to just search the forums).
If you could figure out how to make indexing faster than what exists out there that doesn't require a Google datacenter, you would win a Nobel easily.
Keeping count of words used in a thread would be very fast I suspect, especially with a hash table. Increasing the count of a word is close to like O(1) timing or something.
I think indexing for search requires a shitload more work.
The only bottleneck I see is then iterating through the entire hash table and finding the words with the highest values.
I suspect though, there may be a good way to do that shit. Maybe throw each word entry into a binary heap where the count value is it's weight. This would allow you to instantly pop out the next highest value word in O(1) timing.
But shit, then you'd have to get all the words in ABC order...
Don't be rude in my forum. I don't appreciate it.
We have no plans to implement your suggestion.