Reducing miscateogrized posts

[justusranvier]

If someone was designing new forum software from scratch, one of the things they should do is rethink categorization of posts.

Clearly, many users don't know or are not willing to post topics in the correct section of the forum. This can be caused by simple ignorance or malice. Either way, it greatly distracts from the usability of the forum and sucks up moderator time and attention.

Letting posters the category first and them cleaning up after them later is suboptimal, so why not reverse the process?

A user composes an new thread, but does not get to choose the category up front. When they are ready to submit their post, the forum chooses the category for them by having the poster interact with a finite state machine via a series of questions similar to the following:

1. This post is written in:

• English
• Español
• 中文

2. The primary topic of this post is:

• The Bitcoin network, software, or currency
• A different cryptocurrency
• Something else

The FSM end state is the section where the thread is posted.

Optionally, the forum software could do simple text analysis of the post and thread title to find possible duplicate threads, and ask the user to affirm their thread is new and not a dup.

Lying during the submission process in order to deliberately miscategorize a thread is a bannable offense.

The forum tracks member status (brand new, newbie, full member, hero member, etc) on a per-section basis taking into account how many of their threads get moved to other sections. If a user maintains a sufficiently high accuracy rating they "unlock" that section and can post to it directly without going through the FSM.

This would be a lot of work to implement up front, but I suspect it would greatly reduce the ongoing moderation burden, especially because it removes most or all plausible deniability for trolls.

[SaltySpitoon]

That sounds like a neat feature, hopefully something interesting like that could be implemented with the new forum software?

[railzand]

If 'it' can analyse the text, perhaps it could stop threads that ask the same questions again and again.

[Cyrus]

I think this would fit better in New Forum Software section.

[wtogami]

I like the idea.  wangbus had ideas that might enable things like this, although we haven't thought much about automated categorization yet.

[-ck]

I think this would fit better in New Forum Software section.

That would make this post about miscategorised posts a miscategorised post.

[justusranvier]

The "New forum software" sub-forum did not exist when this thread was created.

Maybe it can lead to a discussion about the best way to handle old threads that fins themselves miscategorized after a new sub-forum has been added.

I think the process of reporting miscategorized threads should be as streamlined as possible, because in the case of countering spammers and trolls the defenders are always outnumbered.

So maybe every thread has a javascript link for reporting a miscategorized post. When clicked, this list could show the path the poster took through the FSM, and the reporter can click on the erroneous decision.

Then they answer the classification questions from that point on and when they arrive at a new end point the report is queued for moderator review and approval.

This process works both for maliciously and retroactively miscategorized posts, but the algorithm should not penalize a miscategorization if the new forum area is newer than the original post.

[activecoin]

I could not find a better place to post this,

I spend alot of time in the altcoins subs and like to go back 10-20 pages to see if there is something I might have missed, starting even to go a little further, I find repeated and almost worthless post's covering the same topic's (not saying they are worthless to who posted them), in the new forum  could it be possible that when typing into the subject line while someone is creating a new topic that a feature like googles search prediction work in a drop down and have it bring up very similar subjects that are already posted to the forum? This could help to consolidate the same questions in the same thread and would even assist user's cause they would be able to see how their question or topic was resolved at an earlier date, it could even help with clutter when a new user might be discouraged from posting more of the same after they see that their question has somewhat been asked before.

If someone was designing new forum software from scratch, one of the things they should do is rethink categorization of posts.

Clearly, many users don't know or are not willing to post topics in the correct section of the forum. This can be caused by simple ignorance or malice. Either way, it greatly distracts from the usability of the forum and sucks up moderator time and attention.

Letting posters the category first and them cleaning up after them later is suboptimal, so why not reverse the process?

A user composes an new thread, but does not get to choose the category up front. When they are ready to submit their post, the forum chooses the category for them by having the poster interact with a finite state machine via a series of questions similar to the following:

1. This post is written in:

• English
• Español
• 中文

2. The primary topic of this post is:

• The Bitcoin network, software, or currency
• A different cryptocurrency
• Something else

The FSM end state is the section where the thread is posted.

Optionally, the forum software could do simple text analysis of the post and thread title to find possible duplicate threads, and ask the user to affirm their thread is new and not a dup.

Lying during the submission process in order to deliberately miscategorize a thread is a bannable offense.

The forum tracks member status (brand new, newbie, full member, hero member, etc) on a per-section basis taking into account how many of their threads get moved to other sections. If a user maintains a sufficiently high accuracy rating they "unlock" that section and can post to it directly without going through the FSM.

This would be a lot of work to implement up front, but I suspect it would greatly reduce the ongoing moderation burden, especially because it removes most or all plausible deniability for trolls.

This is something I dislike about craigslist when I post ads there, although it is a good idea to keep things from being miscategorized often, could it be that this is a tool that could be imposed on new user's for a time period? and then reinstated for another period of time if Mods move a topic of a user that miscategorizes topics 2 or 3 times?

[justusranvier]

This is something I dislike about craigslist when I post ads there, although it is a good idea to keep things from being miscategorized often, could it be that this is a tool that could be imposed on new user's for a time period? and then reinstated for another period of time if Mods move a topic of a user that miscategorizes topics 2 or 3 times?

If you read the proposal's penultimate paragraph you'll see that users earn the ability to skip the FSM for certain sections after they have a track record of properly posting in that section.

The forum tracks member status (brand new, newbie, full member, hero member, etc) on a per-section basis taking into account how many of their threads get moved to other sections. If a user maintains a sufficiently high accuracy rating they "unlock" that section and can post to it directly without going through the FSM.

[activecoin]

If I post to a section accurately I can unlock it? Yet others remain locked and bothersome for a period of time until they unlock? Am I understanding this correctly? As my forum status changes over time by this model all sections unlock? Or is it section by section as I begin to use them?
 

 

[justusranvier]

If I post to a section accurately I can unlock it? Yet others remain locked and bothersome for a period of time until they unlock? Am I understanding this correctly? As my forum status changes over time by this model all sections unlock? Or is it section by section as I begin to use them?

Just because somebody has been very active in the Marketplace sections, for example, does not automatically mean they are knowledgeable about the Mining hierarchy.

In order to post to forum sections which you haven't "unlocked" yet, you go through the series of questions in the FSM to make sure your post ends up being placed in the right area.

After some number of successful posts in a section, presumably you know what kinds of threads are supposed to go in that section.

[kuusj98]

If I post to a section accurately I can unlock it? Yet others remain locked and bothersome for a period of time until they unlock? Am I understanding this correctly? As my forum status changes over time by this model all sections unlock? Or is it section by section as I begin to use them?

Just because somebody has been very active in the Marketplace sections, for example, does not automatically mean they are knowledgeable about the Mining hierarchy.

In order to post to forum sections which you haven't "unlocked" yet, you go through the series of questions in the FSM to make sure your post ends up being placed in the right area.

After some number of successful posts in a section, presumably you know what kinds of threads are supposed to go in that section.

This is for newbie's I suppose, I mean im not going to unlock every single subforum just for the sake of being able to post there, you know.

[jambola2]

Alternatively , and much more feasibly , just have an admin/mod put a flowchart  in the Meta forums.

Most posters who really have something to say try their best to find the appropriate subforum.

A flowchart guide or even more simply , a list of all subforums and a short description would take a short time to make and would save a lot of moderating time.

[Welsh]

That's a really nice feature suggestion, although there may be problems and we will still see posts in the wrong categories this could reduce the work load of the moderators and also reduce the irritation of others. I've never seen such a system be put in to place and I believe we could experiment with it.

It's always a good idea to have a back up feature. This would be extremely useful when moderators/admins are not online and can't move topics immediately. At the moment when reporting a thread it can take a few hours for it to be dealt with because no staff members are online. So this could be used as a fail safe. So having this little wizard may prevent some threads getting posted in something completely unrelated to the topic.
 
I'm not too sure about the being banned for "lying". Because, some people may not understand the question and put a wrong answer. If it's a obvious attempt to miscategorize then there should be warnings given out. For example, the question is "What language is the post written in?" and they respond with German when it's in English. Because, in my eyes this is directly related to trolling.

I like this idea and think this should be a item put up for consideration for the new forum software.

[Cryptopher]

It wouldn't be too difficult to implement a simple categorisation mechanism that uses a representation such as Bag-of-Words. As this would simply capture the unordered occurrence of words, and therefore weak semantics you might only want to use it as a supplementary tool.

The obvious pitfall to using this method is that the dictionary of words i.e. the vocabulary would be pretty large if computed from existing posts, therefore the dimensionality of the feature vector would be significant. The computational trade-off could limit quality.

Furthermore, there is a lot of inter-class similarity of vocabulary between sub-forums in particular.

I think that the original suggestion in the topic is nice, but it could be pretty heavy duty, and it would require manual mapping between potential answers and forums/sub-forms. The addition of some kind of automated categorisation could perhaps indicate a better choice of category.

[hilariousandco]

Uwotm8? Inter-dimensional vocabulary? You sound like Spock or something haha.

It's an interesting idea and people do post a hell of a lot in the wrong sections, but I don't think it's a massive issue. The mods sort it out quickly enough. This sugestion just sounds like way too much work and I can't see how it would work properly in automatically detecting the correct forum without giving false positives half of the time.

[Cryptopher]

The system we have now to correct things manually upon identification is plenty good when you consider it requires no further changes to maintain.

It's not a massive problem anyway.

[mprep]

The system we have now to correct things manually upon identification is plenty good when you consider it requires no further changes to maintain.

It's not a massive problem anyway.

Oh, trust me: sometimes it is. Though it depends on the timeframe, but overall there are lots of posts that I move everyday. Some seem like honest mistakes, while some did that intentionally.

Back on topic, I think that even some rules can be enforced easier this way. For example a question: "Is this thread is about a website?" AND IF ANSWERED YES "Have you already made a thread about it?" would reduce on duplicate threads if the poster just didn't know the rules.

[Cryptopher]

The system we have now to correct things manually upon identification is plenty good when you consider it requires no further changes to maintain.

It's not a massive problem anyway.

Oh, trust me: sometimes it is. Though it depends on the timeframe, but overall there are lots of posts that I move everyday. Some seem like honest mistakes, while some did that intentionally.

Back on topic, I think that even some rules can be enforced easier this way. For example a question: "Is this thread is about a website?" AND IF ANSWERED YES "Have you already made a thread about it?" would reduce on duplicate threads if the poster just didn't know the rules.

I suppose topics aren't posted that often, so to ask a few questions to guide the post to the correct forum wouldn't be too bothersome. The trouble I find is that you can't get a decent mapping between answers and forums without asking lots of questions.

Your proposal to ask questions to tackle the problem of duplicate threads could be useful, however once people are savvy to the mechanism they will simply click no, allowing them to post the topic. Also there may be legitimate cases to make a new topic.

I like the idea of using the questions to tackle other problems, rather than the miscategorisation issue, which is a complex one.

[mprep]

The system we have now to correct things manually upon identification is plenty good when you consider it requires no further changes to maintain.

It's not a massive problem anyway.

Oh, trust me: sometimes it is. Though it depends on the timeframe, but overall there are lots of posts that I move everyday. Some seem like honest mistakes, while some did that intentionally.

Back on topic, I think that even some rules can be enforced easier this way. For example a question: "Is this thread is about a website?" AND IF ANSWERED YES "Have you already made a thread about it?" would reduce on duplicate threads if the poster just didn't know the rules.

I suppose topics aren't posted that often, so to ask a few questions to guide the post to the correct forum wouldn't be too bothersome. The trouble I find is that you can't get a decent mapping between answers and forums without asking lots of questions.

Your proposal to ask questions to tackle the problem of duplicate threads could be useful, however once people are savvy to the mechanism they will simply click no, allowing them to post the topic. Also there may be legitimate cases to make a new topic.

I like the idea of using the questions to tackle other problems, rather than the miscategorisation issue, which is a complex one.

Topics are made quite often, just many of them get deleted or moved fast. Also, if you made a duplicate post with the mechanism in place, one could safely assume this intent was malicious.