Newsvine's Josh Yockey On Grawlix And The Profanity Filter

Does this mean 'no swearing' or does it mean 'no filtering'?
Photo by christopherdale

I counted over fifty grawlix in an article once. I don't think I knew quite the impact and character of the Newsvine profanity filter until then. I didn't see much cursing and when I did it was in the form of the grawlix, a term coined by cartoonist Mort Walker to describe the string of typograpy indicating colorful imprecations.

Thought it was more a Code of Honor thing that people would take time to insert their own marks instead of spelling it out. What manners! But as myself and others soon discover, the grawlix is an automatic replacement string put there by a profanity filter. It's on by default, but you can turn it off. In fact, many in the Newsvine community have expressed that this place is a lot more fun with the filter turned off. It's a preference on your 'Account' tab, and Josh Yockey - Newsvine Director of Technology and Grawlix Development - tells us that the programming token passed out when you turn your filter off is called the "Sailor Token" so as to honor the best of the true cuss craftsmen.

In 1964, Walker (Beetle Bailey, Hi & Lois) invented the term grawlix. The comics artist also contrived plewd, squeen and spurl, thus putting words to the many devices that comic strip artists have used throughout the years. Concocted mainly for amusement, the terms Walker catalogued in an article for the National Cartoonists Society later became The Lexicon of Comicana. The words caught on to some degree to describe drops of sweat, stars about the head, wiggly lines, cursing etc.

Ah, when used well, there's nothing like a well-placed @!$%#. And sometimes there will simply be no better way to describe something, say... talking about a profanity filter. I had a chance to ask Josh a few questions about the grawlixing of Newsvine.

SM: I reckon Carlin's Seven Dirty Words are a good starting point. Can you tell us some of the more unusual words that are filtered?

JY: I've always thought of Carlin's 7 as a little odd, because he includes a couple that are pretty tame. 'Piss' is a little vulgar, but hardly obscene, and we don't filter it out. I don't know what to make of 'tits'. It's far more tame than a few entries that Carlin left out, like '@!$%#' and '@!$%#'.

There's really nothing unusual in the Newsvine list. My previous employer (ESPN) also used a profanity filter, but rather than substrings it only filtered exact matches, so the list had to be comprehensive and by the time I left it had over 300 entries, including the truly bizarre. Reading the entire list in one go made you want to take a shower afterwards.

Approximately how many words are filtered?

A small handful of strings are filtered (less than 10), but they are filtered out even as substrings of larger words, so the potential number of words it cleans includes all of the variations and is pretty comprehensive.

Most words are pretty obvious when determining profanity, but some could fall into a grey area. What are the more contentious words in regards to being on or off the Newsvine list?

I think we err on the side of not filtering as much. Carlin's 'tits' will show up on the site, along with most references to genitalia. We do filter out '@!$%#' even though you can hear it on prime time TV these days.

Is the filter pretty much limited to dirty words, or are there, for example, racial epithets being filtered?

No, it's just dirty words. The reason for that is that profanity is a matter of taste, and for the people who decide that they enjoy using 'sentence enhancers', we have no problem with it - within reason - as long as other people can opt out. (Or in this case, refuse to opt in.) Racial epithets, on the other hand, are not a matter of taste and they will get your comment deleted and the author likely banned.

Not all words can be anticipated. Does the list of filtered words ever get tweaked?

We haven't changed the profanity list since we wrote it, because filtering substrings tends to automatically take care of 'creative' profane neologisms. On the other hand we have a list of exceptions, because substring filtering can be a little too aggressive. For example, Matsushita was added after that company had been in the news and we noticed the overgrawlixing.

Maybe you could expound a little bit on the social and cultural aspects of such a tool. Are there any non-English words that are filtered?

No. We (reluctantly) made the decision early on that Newsvine would have to be an English-only site. All of the technology supports internationalization (and the first few weeks after launch we were pleased to see articles in Arabic, German, Japanese, and Chinese), but we don't have the resources to moderate discussions in every language. So our profanity filter stays English-only as well.

It's surprisingly non-controversial. I can't think of any case where a user was confused about what had happened to their comment and why. It seems profanity filtration and grawlixes are well-established enough in our culture that everyone gets it.

Are there any places where the profanity filter does not work?

If you mean text areas on the site where the filter isn't being applied, that's a bug. A few of them are known (e.g. Newsvine Live) and we just haven't come up with a good way to interject the filter. But by design, all of the text should run through the filter for consistency.

How was the character order and overall aesthetic appearance arrived at of this particular Newsvine grawlix?

I came up with the string on the fly. It's hardcoded and has never changed. The filter is a very early feature and was in place before the end of the closed beta. I think my aesthetic was to use larger, blocky symbols so it stands out as a dense rectangle of nonsense on a quick scan.

Did anyone know there was a name for that string of characters when Newsvine developed their profanity filter?

I don't think anyone in the office did. We referred to it as the garble. My spellchecker just highlighted grawlix as misspelled and recommends garlic as the replacement.

Back To Top