=][= WORD FILTER EXPERIMENTATION ONGOING =][=

[Brother Tyler]

We are looking at expanding our word filter entries.

 

Currently, the various filters we have convert some string of text into some other string of text. Most of the time this means turning profanity into , but there are also times where we redirect IP-violating websites to the Games Workshop site or correct common misspellings (e.g., there are a few variations of "emperor" that are covered).

 

The software update a few years ago gave us two additional options that we haven't implemented, but we are looking at ways they might be beneficial to the community (and some of our existing word filters might be converted to one of these new methods).

 

The first new method blocks certain text strings from being submitted - if you try to submit a post with one of these text strings, the software will tell you that you can't and give you the opportunity to change it to something acceptable.

 

The second new method is to require staff review and approval before content with the filtered string is published - if you submit content with one of these text strings, the content will remain hidden. You will be notified of the requirement for staff review and approval, and will be notified when the content is approved. We haven't tested out content that isn't approved yet, but I expect that you'll receive some form of notification if that happens, too.

 

We are currently only experimenting with the second new method, and have selected a half dozen or so text strings to be filtered. These are words/phrases/acronyms that are generally off-topic here, but which occasionally crop up and usually signal that a discussion is getting out of hand. Sometimes, however, these words/phrases/acronyms have legitimate and on-topic uses, hence the option for us to approve them.

 

No, I'm not going to tell you what the experimental phrases are. I selected a few from across the spectrum of viewpoints, but those might not be in our final list (assuming we decide to implement either/both of these methods).

 

I'm just letting you know so that you'll understand what's going on if you get a notification about your content requiring staff review and approval.

 

Our intent is to not go overboard, but to have a reasonable selection of filters that we can adjudicate fairly, hoping to better filter potentially objectionable content and keep everyone's passions from getting out of control (i.e., keeping the community a welcoming place for everyone).

 

Now if you have any suggestions for words/phrases/acronyms that you think should be filtered, please provide those suggestions to me via PM (we don't need to get into a public donnybrook over these ideas). The staff will discuss suggestions and determine which, if any, should be included in our filtering.

[Brother Tyler]

Don't be surprised if you submit content and receive a message that it requires moderator approval.

 

The new software (okay, we've had it for a couple of years now) has expanded the options we have for word filters.

 

The old system simply replaced a text string with some other text string, and we used that to either change profanity to the  emoji, correct a few common misspellings (you would be amazed at the variations we saw for "Emperor"), and automatically enter a few diacritical marks (e.g., you can type in Khârn and Ulthwé and the system automatically changes them to have the marks).

 

The new system preserves that option, but gives us two other options. One of those simply blocks the content from being submitted if certain text strings are included. The other requires that a moderator review and approve the content if it contains certain text strings. We expanded our spelling fixes (i.e., we added Slaanesh and Tzeentch), but our main expansion has been adding words/acronyms/phrases that will require moderator approval. In most cases, these are words/acronyms/phrases that might be appropriate/acceptable (hence the option for them to be approved), but they might also be inappropriate/unacceptable or otherwise signal that a discussion is starting to go off the rails (e.g., Godwin's law and other similar issues).

 

So if you include a word/acronym/phrase that is included in this filtering option, you'll receive a message that the content must be approved by a moderator. The moderators and administrators will receive a notification and one of us will review the content. If it is used in a manner that is appropriate for our community, the content will be approved and you'll receive a notification. If the content is not used in a manner that is appropriate for our community, it will not be approved and you should receive an explanation.

 

This is all very experimental right now and we don't even know if the words/acronyms/phrases that we currently have marked for this are the ones that we'll go with (and it seems likely that whatever words/acronyms/phrases we select, we'll probably have to expand on them over time). In a nutshell, we're focusing on those words/acronyms/phrases that are off-topic and/or offensive, and while we want to cover all the bases, we don't want to stifle relevant discussion (so we're trying not to go overboard). The review/approval method is good for this (instead of replacing the text string or blocking the content).

 

Once we're satisfied with things, including the staff's standards, processes, and procedures for approving/disapproving, we'll provide a more in-depth explanation. We have already identified some problematic filters, so this experimentation phase is helping us to fine-tune things and ensure that we're not being too heavy-handed so that the end result promotes constructive and respectful discourse within the community.

 

We are open to suggestions on this, especially if you have suggestions for things that might be included or if you wonder why your content was filtered in this way.

[Moonstalker]

Didn't know there was already a word filter desu

[Xirix]

I feel like an auto-correct for some spellings is good and bad, the good is obvious, but I feel it's not going to help people actually learn to correctly spell whatever is being corrected when they can just continue to spell it wrong and be rewarded.

[Marshal Rohr]

4 minutes ago, Xirix said:

I feel like an auto-correct for some spellings is good and bad, the good is obvious, but I feel it's not going to help people actually learn to correctly spell whatever is being corrected when they can just continue to spell it wrong and be rewarded.

The Bolter and Chainsword: the internet’s premier warhammer 40,000 forum and grammar school!

[BrainFireBob]

Best of luck to the mods, but beyond profanity, this will likely descend into river crab/water horse territory, in my experience

Edited October 18 by BrainFireBob

[Mendi Warrior]

Thanks for the clear communication on this and good luck.

 

This reminded me about this: Hypersensitive Profanity Filter Censors ‘Bone’ at Paleontology Conference The article contains a link to a list of words banned by the algorithm, my understanding being that most can have various meanings and some are quite central to paleontology. I haven't looked at the list yet.

 

Excerpt from the article: 

Quote

“Most funny to us was the censorship of ‘bone,’ which, after all, are the main thing we work with,” Holtz tells the Times.

 

PS: we'll see if this post generates a warning 

[Gillyfish]

To be clear, the word filter doesn't generate warnings - that still requires mod intervention and some judgement about the non-Board appropriate language being used (we're like Father Christmas, we can tell, you know... ).

 

As you can tell, bone is not in the filter because it is used pretty commonly in the hobby. However, it should not be used inappropriately.

[Emperor Ming]

It should filter everything we say into high gothic

[Marshal Rohr]

The filter just grabbed my post with no swearing, innuendos, jokes, or mischief in the Dark Mech thread so some fine tuning may be needed. 

[Brother Tyler]

On 10/18/2024 at 12:17 PM, Marshal Rohr said:

The filter just grabbed my post with no swearing, innuendos, jokes, or mischief in the Dark Mech thread so some fine tuning may be needed. 

 

Actually, the filter caught a text string that it was supposed to catch; and since that text string was used in an acceptable manner, the post was approved post haste. This is an instance demonstrating the effectiveness of the filter.

 

Now whether or not that text string remains in the filter is up for debate. It was part of a broad range of text strings that were added to catch potentially off-topic or offensive political discussion. The question at this point is the likelihood of that text string (and several similar text strings) actually coming up here in the unacceptable manner compared to your acceptable usage.

[Nova-V]

As someone whose work has in the past lead to making tables for exclusion/inclusion of words it's a thankless and nuanced job.  People just don't seem to understand how literal most text censors are and seem to expect it to just work by intuition (people in the world of UK addresses know the joy of an over zealos censor banning all mail to a town called S:cuss:horpe).  Really nice to see a community taking the time to properly improve a censor system and use a human pair of eyes when needed despite the extra workload.

[Urauloth]

I realise this is an edge case, but I wonder if it's possible to get some exception for the word "Nigh:cussch" because the censor can be a little tedious on that one if you're trying to discuss the Thramas Crusade.

[Brother Tyler]

That's a good catch. I've changed the filter so that the text string must be reviewed and approved, rather than being replaced. I've also temporarily added a filter that will [hopefully] fix all of the existing places where the old filter changed things.