3 ms·

Show HN: A naive classifier to figure out if a sentence contains dirty words

10 points by gauravm 11y ago

kristiandupont 11y ago
A friend of mine told me about how they once had to dig through their code to figure out why their site was classified as adult by some filter. After days of searching, they found this comment at the bottom of a javascript file: // Slut. Which is Danish for "the end".
- renekooi 11y ago
  Another false positive anecdote! My parents used to have their ISP's adult content filter enabled. One day I couldn't visit DeviantArt, because it said "Mature Content Filter Enabled" somewhere on the page, and "mature content" triggered the ISP's filter.
nopuremore 11y ago
With the little effort of google translate your dirty words to Spanish (copy paste all words), you obtain a filter for Spanish, add synonyms for stronger filtering. Perhaps gay is not a dirty word? (is included in your dirty words, but gay people should think otherwise.
- spoiler 11y ago
  I'm gay, but I don't consider it offensive that the word is in there. A lot of people use the term "gay" in conversation as a synonym for "that sucks"; a friend of mine does it all the time. I don't think they mean anything by it. To differentiate between "I am gay," and "Oh that's gay. I'm sorry that happened," you'd need a NLP with a politeness preference.
  - gauravm 11y ago
    Sorry, no offense intended, if anyone took it. In my use-case, the words such as 'gay' and 'lesbian' were in almost all cases, used for explicit documents. This is a very naive implementation to quickly get a handle of amount of porny documents. I intend to do some more work around clustering of porny words. I think understanding sentiment would be hard and involves a lot of labeled data, but that is a potentially very useful project.
    - spoiler 11y ago
      It's okay! I wasn't offended. :-) Although I didn't realise this was meant to filter out a pornographic vocabulary; it makes more sense now.
spoiler 11y ago
I don't find this very useful. It's too naïve for a real-world usecase. I didn't look at the implementation, but the "classy party" looks like it simply matches for a sequence of 'a', 's', and 's' bytes in a string. It would be better it it tokenized the sentence using punctuation and white-space as terminators. So, it would detect `big-ass sandwich` and `smart-ass person` but not `classy party` or `bass instrument`. Furthermore, it would be cool if you created a configuration format for this kind of thing, so one could do something like this (excuse the config format, I realise it's probably shit and problematic): [smart][big][fat]ass !sex[ual]+education which would detect all of the following: smartass, bigass, fatass, and ass itself. The second rule would not filter `sex(?:ual)` token followed by an `education` token. You get the idea These are just some heat-of-the-moment ideas, because I think this is exciting and could be useful. :-)
- gauravm 11y ago
  Thanks. This quick idea worked for my cases, because there were few potential false positives. But your idea around using a regex style matcher should be good.
natch 11y ago
What is the use case for such a classifier?
- gauravm 11y ago
  I wanted something easy to use to quickly get an idea of how much explicit content could we be dealing with. The main challenge was dealing with a multi-lingual database. I didn't even find a naive classifier. Though I don't have time/RoI to improve this, but potential ideas are to use labeled data to cluster porny words and get a probablistic metric of porni-ness of a sentence.
radio4fan 11y ago
I inherited a (dreadful) application which had a hilariously lame 'rude words filter'. It checked for words on a banned list. The full list is here: http://pastebin.com/raw.php?i=1Pv4v8j7 http://pastebin.com/raw.php?i=1Pv4v8j7 It contains such gems as "cockburger", "penispuffer", and -- the piece de resistance -- "twatwaffleunclefucker".

10 points by gauravm 11y ago

A friend of mine told me about how they once had to dig through their code to figure out why their site was classified as adult by some filter. After days of searching, they found this comment at the bottom of a javascript file: // Slut. Which is Danish for "the end".

renekooi 11y ago
Another false positive anecdote! My parents used to have their ISP's adult content filter enabled. One day I couldn't visit DeviantArt, because it said "Mature Content Filter Enabled" somewhere on the page, and "mature content" triggered the ISP's filter.

nopuremore 11y ago

With the little effort of google translate your dirty words to Spanish (copy paste all words), you obtain a filter for Spanish, add synonyms for stronger filtering. Perhaps gay is not a dirty word? (is included in your dirty words, but gay people should think otherwise.

spoiler 11y ago
I'm gay, but I don't consider it offensive that the word is in there. A lot of people use the term "gay" in conversation as a synonym for "that sucks"; a friend of mine does it all the time. I don't think they mean anything by it. To differentiate between "I am gay," and "Oh that's gay. I'm sorry that happened," you'd need a NLP with a politeness preference.
- gauravm 11y ago
  Sorry, no offense intended, if anyone took it. In my use-case, the words such as 'gay' and 'lesbian' were in almost all cases, used for explicit documents. This is a very naive implementation to quickly get a handle of amount of porny documents. I intend to do some more work around clustering of porny words. I think understanding sentiment would be hard and involves a lot of labeled data, but that is a potentially very useful project.
  - spoiler 11y ago
    It's okay! I wasn't offended. :-) Although I didn't realise this was meant to filter out a pornographic vocabulary; it makes more sense now.

spoiler 11y ago

I don't find this very useful. It's too naïve for a real-world usecase. I didn't look at the implementation, but the "classy party" looks like it simply matches for a sequence of 'a', 's', and 's' bytes in a string. It would be better it it tokenized the sentence using punctuation and white-space as terminators. So, it would detect `big-ass sandwich` and `smart-ass person` but not `classy party` or `bass instrument`. Furthermore, it would be cool if you created a configuration format for this kind of thing, so one could do something like this (excuse the config format, I realise it's probably shit and problematic): [smart][big][fat]ass !sex[ual]+education which would detect all of the following: smartass, bigass, fatass, and ass itself. The second rule would not filter `sex(?:ual)` token followed by an `education` token. You get the idea These are just some heat-of-the-moment ideas, because I think this is exciting and could be useful. :-)

gauravm 11y ago
Thanks. This quick idea worked for my cases, because there were few potential false positives. But your idea around using a regex style matcher should be good.

natch 11y ago

What is the use case for such a classifier?

gauravm 11y ago
I wanted something easy to use to quickly get an idea of how much explicit content could we be dealing with. The main challenge was dealing with a multi-lingual database. I didn't even find a naive classifier. Though I don't have time/RoI to improve this, but potential ideas are to use labeled data to cluster porny words and get a probablistic metric of porni-ness of a sentence.

radio4fan 11y ago

I inherited a (dreadful) application which had a hilariously lame 'rude words filter'. It checked for words on a banned list. The full list is here: http://pastebin.com/raw.php?i=1Pv4v8j7 http://pastebin.com/raw.php?i=1Pv4v8j7 It contains such gems as "cockburger", "penispuffer", and -- the piece de resistance -- "twatwaffleunclefucker".